Full Machine Learning Week 7-Track Agenda 2021 – Detailed Session Descriptions
Predictive Analytics World
May 24-28, 2021
See the full 7-track agenda for the six co-located conferences at Machine Learning Week. A Machine Learning Week Ticket is required for full access. To view the agenda for one individual conference, click here: PAW Business, PAW Financial, PAW Industry 4.0, PAW Climate, PAW Healthcare, or Deep Learning World.
Session Levels:
Blue circle sessions are for All Levels
Red triangle sessions are Expert/Practitioner Level
Machine Learning Week - Virtual - Day 1 - Monday, May 24th, 2021
Amazon's vision is to be earth's most customer-centric company. This talk explores how the Alexa Hybrid Science team in Pittsburgh, PA applies a customer-centric lens to cutting-edge machine learning research. The team is responsible for developing on-device Alexa automatic speech recognition models to provide a faster, more reliable Alexa experience. Our research includes neural network compression techniques, end-to-end spoken language understanding and optimizing machine learning for edge devices.
Amazon's vision is to be earth's most customer-centric company. This talk explores how the Alexa Hybrid Science team in Pittsburgh, PA applies a customer-centric lens to cutting-edge machine learning research. The team is responsible for developing on-device Alexa automatic speech recognition models to provide a faster, more reliable Alexa experience. Our research includes neural network compression techniques, end-to-end spoken language understanding and optimizing machine learning for edge devices.
Amazon's vision is to be earth's most customer-centric company. This talk explores how the Alexa Hybrid Science team in Pittsburgh, PA applies a customer-centric lens to cutting-edge machine learning research. The team is responsible for developing on-device Alexa automatic speech recognition models to provide a faster, more reliable Alexa experience. Our research includes neural network compression techniques, end-to-end spoken language understanding and optimizing machine learning for edge devices.
Amazon's vision is to be earth's most customer-centric company. This talk explores how the Alexa Hybrid Science team in Pittsburgh, PA applies a customer-centric lens to cutting-edge machine learning research. The team is responsible for developing on-device Alexa automatic speech recognition models to provide a faster, more reliable Alexa experience. Our research includes neural network compression techniques, end-to-end spoken language understanding and optimizing machine learning for edge devices.
Amazon's vision is to be earth's most customer-centric company. This talk explores how the Alexa Hybrid Science team in Pittsburgh, PA applies a customer-centric lens to cutting-edge machine learning research. The team is responsible for developing on-device Alexa automatic speech recognition models to provide a faster, more reliable Alexa experience. Our research includes neural network compression techniques, end-to-end spoken language understanding and optimizing machine learning for edge devices
Amazon's vision is to be earth's most customer-centric company. This talk explores how the Alexa Hybrid Science team in Pittsburgh, PA applies a customer-centric lens to cutting-edge machine learning research. The team is responsible for developing on-device Alexa automatic speech recognition models to provide a faster, more reliable Alexa experience. Our research includes neural network compression techniques, end-to-end spoken language understanding and optimizing machine learning for edge devices.
The quality of any machine learning or deep learning model depends on the values that define the model structure and corresponding hyperparameters. Many practitioners may find themselves investing countless hours manually searching for the right model and related hyperparameter values. Some use highly inefficient grid search methods. Others will use simple random sampling, which actually works fairly well. But alone, this method only offers a globalized search, and other sampling methods may be better suited to the job.
Why not use machine learning to automate the search for the best model?
This presentation details an advanced approach that uses both global and local search strategies that can be evaluated in parallel to ensure a quick and efficient exploration of the decision space. In the case of this presentation, a geneticalgorithm (GA) will be examined for the global search because the selection and crossover aspects of a GA distinguish it from a purely random search. A generating set search (GSS) will be used to greedily search the local decision space.
The quality of any machine learning or deep learning model depends on the values that define the model structure and corresponding hyperparameters. Many practitioners may find themselves investing countless hours manually searching for the right model and related hyperparameter values. Some use highly inefficient grid search methods. Others will use simple random sampling, which actually works fairly well. But alone, this method only offers a globalized search, and other sampling methods may be better suited to the job.
Why not use machine learning to automate the search for the best model?
This presentation details an advanced approach that uses both global and local search strategies that can be evaluated in parallel to ensure a quick and efficient exploration of the decision space. In the case of this presentation, a genetic algorithm (GA) will be examined for the global search because the selection and crossover aspects of a GA distinguish it from a purely random search. A generating set search (GSS) will be used to greedily search the local decision space.
The quality of any machine learning or deep learning model depends on the values that define the model structure and corresponding hyperparameters. Many practitioners may find themselves investing countless hours manually searching for the right model and related hyperparameter values. Some use highly inefficient grid search methods. Others will use simple random sampling, which actually works fairly well. But alone, this method only offers a globalized search, and other sampling methods may be better suited to the job.
Why not use machine learning to automate the search for the best model?
This presentation details an advanced approach that uses both global and local search strategies that can be evaluated in parallel to ensure a quick and efficient exploration of the decision space. In the case of this presentation, a genetic algorithm (GA) will be examined for the global search because the selection and crossover aspects of a GA distinguish it from a purely random search. A generating set search (GSS) will be used to greedily search the local decision space.
The quality of any machine learning or deep learning model depends on the values that define the model structure and corresponding hyperparameters. Many practitioners may find themselves investing countless hours manually searching for the right model and related hyperparameter values. Some use highly inefficient grid search methods. Others will use simple random sampling, which actually works fairly well. But alone, this method only offers a globalized search, and other sampling methods may be better suited to the job.
Why not use machine learning to automate the search for the best model?
This presentation details an advanced approach that uses both global and local search strategies that can be evaluated in parallel to ensure a quick and efficient exploration of the decision space. In the case of this presentation, a genetic algorithm (GA) will be examined for the global search because the selection and crossover aspects of a GA distinguish it from a purely random search. A generating set search (GSS) will be used to greedily search the local decision space.
The quality of any machine learning or deep learning model depends on the values that define the model structure and corresponding hyperparameters. Many practitioners may find themselves investing countless hours manually searching for the right model and related hyperparameter values. Some use highly inefficient grid search methods. Others will use simple random sampling, which actually works fairly well. But alone, this method only offers a globalized search, and other sampling methods may be better suited to the job.
Why not use machine learning to automate the search for the best model?
This presentation details an advanced approach that uses both global and local search strategies that can be evaluated in parallel to ensure a quick and efficient exploration of the decision space. In the case of this presentation, a genetic algorithm (GA) will be examined for the global search because the selection and crossover aspects of a GA distinguish it from a purely random search. A generating set search (GSS) will be used to greedily search the local decision space.
The quality of any machine learning or deep learning model depends on the values that define the model structure and corresponding hyperparameters. Many practitioners may find themselves investing countless hours manually searching for the right model and related hyperparameter values. Some use highly inefficient grid search methods. Others will use simple random sampling, which actually works fairly well. But alone, this method only offers a globalized search, and other sampling methods may be better suited to the job.
Why not use machine learning to automate the search for the best model?
This presentation details an advanced approach that uses both global and local search strategies that can be evaluated in parallel to ensure a quick and efficient exploration of the decision space. In the case of this presentation, a genetic algorithm (GA) will be examined for the global search because the selection and crossover aspects of a GA distinguish it from a purely random search. A generating set search (GSS) will be used to greedily search the local decision space.
As companies continue to increase their investment in Machine Learning (ML) and Artificial Intelligence (AI), the focus is increasingly on how to make machine learning pervasive, creating business value by embedding it throughout a company's operations. But what does it take to get out of Proof of Concept purgatory and go from pilots to operational deployments? Do we need more technology, new roles, a different approach or what?
In this session, James will kick off the operational track with a discussion of what it takes to get value at scale from ML. He'll talk about critical issues and how to address them. He'll outline a proven approach that will let non-digital natives and even the most risk-averse companies operationalize and get value from their machine learning investments. Whatever your motivation for using machine learning and no matter how mature (or not) your machine learning team is, you'll learn how to succeed.
Traditionally, successful predictive analytics initiatives have relied on the application of complex algorithms to a lot of numbers and interpreting the output into meaningful language that clients can easily understand. This presentation aims to turn that premise on its head. Starting with a lot of words provided by consumer panelists, NCP analyzed the text provided in their feedback to us and translated those words into numbers. These numbers added up to value for National Consumer Panel. This presentation will share lessons learned from combining text analytics insights with more traditional approaches leveraging quantitative analytics. Attendees can expect to learn how NCP's experience may be extrapolated to provide insights on other business questions.
Healthcare risk adjustment is the most impactful financial mechanism to account for the differences in risk in members that enroll with different carriers. In the ACA program alone, billions of dollars are moved each year between health plan carriers. The calculations rely on risk assessment programs which are typically based on linear regression. Simplicity, understandability, and transparency are paramount considerations in the modeling due to the zero-sum nature of risk adjustment. This brings us to an important question: if healthcare expenditures by member are non-linear (which in fact they are), is the additional complexity of machine learning worth it to model these more accurately?
This session provides an overview of this mathematics that is critical to healthcare financials in each line of business (commercial, medicare, medicaid) and also compares and contrasts traditional methods such as linear regression against machine learning variants that involve use of neural networks. We will also discuss the considerations involved in picking the right tools for a couple of applications related to risk assessment.
A single doctor, overworked and overwhelmed, will make a less-informed decision than a team of doctors. Now, imagine the potential knowledge we could gain if we combined insight from doctors all over the world. In this case study, we will explain how Innovative Precision Health (IPH) uses large amounts of aggregated anonymous population data to learn about the trajectory of a disease and predict the most effective treatment for a patient. By exponentially increasing the amount of information available, IPH is able to provide better predictions for doctors, better care for patients, and profitable outcomes for insurance and pharmaceutical companies alike.
I4.0 has created a plethora of applications and vendors competing for attention in the manufacturing space. Many vendors are pushing platforms and enablement yet what manufacturers really need are hard core scalable use cases that deliver measurable impact and value. This cannot be solved by an individual vendor or platform. I4.0 success is dependent on cooperation and partnership between different players in the I4.0 landscape and a new approach to continuous improvement in order to massively reduce the time it takes to derive action, impact and value from data.
Accurate, timely estimation of carbon emissions is critical for businesses and governments to take action on climate change. The Climate TRACE coalition aims to furnish these estimates for all major sources of emissions globally, on a near real-time basis. In this session we’ll give an overview of our work on estimating emissions from one of the most important sources of greenhouse gases, power generation from coal. The approach focuses upon detecting plumes using a variety of approaches, including multi instance deep learning.
Each real-world application of machine learning owns a set of defining characteristics that distinguish it from other problems. In Nauto's case, we discovered that our driving data is remarkably uniform. By leveraging this trait, we were able to achieve massive optimizations in model accuracy and inference speed, directly improving Nauto's core safety offerings like Forward Collision Warning.
This talk will explore the importance of identifying domain-specific traits when deploying AI solutions to the world. It will show how leveraging the quirks of your problem can lead to optimizations that occasionally even contradict established conventions and common intuition.
Learn how Barron’s Group is leveraging automation and Machine Learning to produce direct-to-reader content and power dynamic editorial tools. This talk will focus on both the technical infrastructure necessary to automate content, and the innumerable product and editorial decisions made throughout the design, development, and deployment phases.
Over the past year, Barron’s Group has used Wordsmith to manage the text generation component of its automated story systems. Auto-generated stories are templated to summarize relevant stock-market events, and may be triggered within our own system to publish under any conditions specified. Because templates can be designed for any relevant use-case, this system is easily extended to different story topics, languages, and business segments. Barron’s currently produces stories on the state of the market at company close, and plans to expand to publish japanese-language translations, automated flash headlines, and morning stock summaries.
Barron’s Group is also in the process of building a Machine Learning-powered system for editorial assists. This tool will rely on anomaly detection performed with Amazon Sagemaker, and will alert editors when noteworthy events occur. In the short-term, this tool is intended as a time-saver for editors who would otherwise have to manually sift through price and volume data for dozens of stocks. In the long term, this kind of technology could be combined with Wordsmith’s templating functionality to produce entire headlines or articles to send off to editors, for review or publication. Here, we will focus primarily on our current process for automated article publication with Wordsmith, along with our work to this point on the Sagemaker editorial assist tool. We will review the results of our work, and discuss future steps that can be taken to expand upon these systems.
With COVID-19 shifting customer behavior the demand for Marriott’s Home and Villas offering has exploded. In this session we will share how deep neural network based Natural Language Generation techniques were used to support Marriott Homes and Villas generating property titles automatically. We will explore different techniques for title generation, provide performance comparisons and lessons learned to activate the model.
Hardly a week goes by without news of money laundering affecting this or that financial institution. As gatekeepers of the financial system, banks play a crucial role in finding and reporting instances of financial crime. In this talk, we describe the challenges of detecting money laundering and outline why employing machine learning techniques is critically important. We will discuss some of the challenges that we faced at ABN AMRO as well as the solutions we came up with, including a method to deal with changes in feature space while running models in production and an in-house-developed model explainability technique.
This session details for big data collected on over 1.4 million people has been used to develop AI-led, personalised interventions that are clinically-proven to reverse type 2 diabetes, which has redefined the use of digital therapeutics in the USA, Canada, United Kingdom, Germany and India.
Scanning Electron Microscope (SEM) images are the primary means used by expert engineers to identify and diagnose defects in semiconductor variable shape beam (VSB) mask writers. Deep learning (DL) offers an attractive alternative to this tedious process. However, extremely robust mask writers preclude collecting a large variety of SEM images to train DL models. Using digital twins that can mimic SEM images provides an exceptional way to synthesize ample DL training data. This talk will take a deep dive into synthesizing SEM images and leveraging them to build DL models for VSB mask writer defects analysis.
The energy landscape is going through a drastic transformation. We are moving away from centralized power plants to more distributed energy resources such as solar, electric vehicles. This transformation makes it very difficult for the grid to handle and manage. Through this session we would explore how using data science and artificial intelligence, industries can optimize their energy usage while aiming to reduce costs and meet their sustainability goals. We'd be delving deeper into some of the forecasting and predictive analytics techniques we have created and measuring their impact. Also, we would share insights from real-life use cases of our customers and how AI has helped them monitor, control, and optimize their energy assets.
Scanning Electron Microscope (SEM) images are the primary means used by expert engineers to identify and diagnose defects in semiconductor variable shape beam (VSB) mask writers. Deep learning (DL) offers an attractive alternative to this tedious process. However, extremely robust mask writers preclude collecting a large variety of SEM images to train DL models. Using digital twins that can mimic SEM images provides an exceptional way to synthesize ample DL training data. This talk will take a deep dive into synthesizing SEM images and leveraging them to build DL models for VSB mask writer defects analysis.
Engineering design simulation generates enormous amounts of structure data, the bulk of which is never studied or mined for insights. Several case studies will be presented for advancing the belief that simulation data presents an opportunity to allow innovation, hidden insights and guidance to emerge thru the use of simple ML techniques.
QuantumScape is developing solid-state lithium metal anode batteries to make long-range, mass market electric vehicles faster charging, longer range, and more affordable. Software has been a key enabler of rapid learning that has spanned new materials discovery through production quality control. QuantumScape has deployed many different models of varying complexity up to neural networks for image segmentation. Working under the constraints of smaller data (10s to 1000s of labels) in a small team of developers has taught the value of empowering expert users to label data and construct their own models.
The increasing prevalence of high-fidelity industrial digital twins is providing a range of opportunities to apply Reinforcement Learning (RL) techniques outside of traditional academic examples and video games. While this trend is now well-established, most RL developments and deployments in the real world are done on an ad-hoc basis with little consideration given to how to repeat and scale similar initiatives in an efficient way. In this session we will address these shortcomings and illustrate them through our experience in optimising the design of a state-of-the-art sailing boat for a prominent competition with RL. Building an agent to control the boat is a very complex RL task for several reasons: imperfect information, loosely defined goals with delayed rewards, highly dynamic state and action spaces. In racing conditions, it takes a team of Olympic-level athletes to sail the boat and make it “fly” thanks to its underwater wings (read hydro-foiling). In order to control convergence variability and sampling efficiency, the working solution required a custom deep learning implementation of the Soft Actor Critic RL algorithm, with state of the art improvements such as experience replay buffer pooling, domain randomisation and curriculum learning. Beyond describing solutions to traditional RL considerations, we will also focus on the underlying workflows and technology stack required to carry out a deep learning project of this technical complexity in a scalable way. We will use facets of Software 2.0, such as higher-level APIs and the automation of end-to-end model development tasks, to highlight our iterative choices and the optimisation opportunities along the machine learning pipeline and ultimately the production system.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
.
Machine Learning Week - Virtual - Day 2 - Tuesday, May 25th, 2021
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Ethics in AI is no longer optional, and profit and human values are not mutually exclusive. As the scale, scope and speed of AI adoption increase, we are seeing more examples of algorithmic discrimination, automated racism, ageism, deepfakes, large-scale manipulation ("nudging"), and other harms. Much of this is unintentional, but were you to deploy such an app, it could severely damage your organization's reputation and top or bottomline. Because no matter what business you are in, we are all in the business of trust. Join this session to lean how to adopt and scale AI ethically and responsibly.
A significant amount of research is conducted on the distribution of talent. Talent dispersion and clustering in different industries such as software engineering is a well-known phenomenon and is discussed widely in academic literature and public media. This talk explores how the pandemic has affected the dispersion of talent and how companies are addressing these changes in the post-vaccine world.
Medicine is an ever-unfolding quest to ensure patients receive life improving therapies to return to full life. Advances in intelligent data, AI, data automation, machine learning, and computational capability allow an efficient pursuit of better outcomes for patients while reducing health care costs. To help understand new approaches to healthcare and new forms of innovation, Maneesh Shrivastav PhD, Director of Market Development, Science and Analytics at Medtronic, will run through innovation at the medical technology company and provide examples of how the company is leveraging data science to improve patients’ lives.
With the need to cater to a global audience, there is a growing demand for applications to support speech identification/translation/transliteration from one language to another. This talk starts off with a brief introduction to the topic of Machine Translation (MT), the evolution/application of Machine Translation. The focus will be on Neural Machine Translation (NMT). Then it moves on to introduce some of the typical customer cases and finally ends with how to embed such functionality in your application. The talk will showcase how maintenance and company records across various subsidiaries which might be in their local language can be standardized into a common language so that all the global data can be used for NLP models.
In applying AI to fight climate change, we appear to have the irresistible force meeting the immovable object. In this talk, I'll present lessons that I've learned from work at Google where both AI and large-scale computation can be used for both climate mitigation and climate adaptation. The talk will include discussing carbon-aware computing, flood forecasting, and using machine learning to accelerate fluid modeling.
In the field of Speech Recognition, the state of the art for generic conversations has reached superhuman levels. However, things are not nearly as good in specialized knowledge domains: conversation in which people with accents speak in a noisy environment often results in high error rates. Considering the low performance of Speech Recognizers on real data, it becomes imperative to customize the end-to-end probabilistic model. This session will focus on discussing the fundamentals of speech recognition and how Cisco has moved from multiple phoneme-based models to a single end-to-end grapheme-based Recurrent Neural Network architecture that can transcribe audios directly while also reducing latency by tweaking the model during inference time.
Predictive analytics reaches out into more and more areas of business, industrial, and research applications. The sheer number of different algorithms and technologies is staggering. In this presentation we give a top-level review of the most popular tree-based algorithms now easily available to anyone through Minitab. From individual trees to powerful modern ensembles, we highlight their strengths, weaknesses, uses, and limitations, especially when compared to the conventional modeling techniques like multiple linear and logistic regression. We illustrate the flexibility of the modern tree-based predictive analytics by building a series of models to predict power generation of a solar energy power plant. In this case, gradient boosting ensemble achieves 20% more accuracy compared to the conventional regression. In addition, the winning model provides great insights into the nature of multivariate dependencies.
Predictive analytics reaches out into more and more areas of business, industrial, and research applications. The sheer number of different algorithms and technologies is staggering. In this presentation we give a top-level review of the most popular tree-based algorithms now easily available to anyone through Minitab. From individual trees to powerful modern ensembles, we highlight their strengths, weaknesses, uses, and limitations, especially when compared to the conventional modeling techniques like multiple linear and logistic regression. We illustrate the flexibility of the modern tree-based predictive analytics by building a series of models to predict power generation of a solar energy power plant. In this case, gradient boosting ensemble achieves 20% more accuracy compared to the conventional regression. In addition, the winning model provides great insights into the nature of multivariate dependencies.
Constructing predictive models using healthcare claims data often requires complex feature engineering and extensive clinical domain knowledge, a tedious, time-consuming and error-prone process. To address this challenge and enable rapid model production, Geneia data scientists developed an automated pipeline to construct machine learning models with little or no manual intervention.
In the automated pipeline, the diagnosis and medication data in raw claims were aggregated into clinically meaningful ‘groupers’, and then the ‘groupers’ were fit into the pipeline for automated model construction. Using this approach, Geneia data scientists were able to speed up the model construction process by around 100 times, while maintaining high accuracy and good interpretability.
Geneia data scientist Zhipeng Liu will discuss the creation of the automated pipeline and an application, a series of models for predicting the onset of major chronic diseases.
Due diligence prior to model deployment involves identifying bias risks, features associated with group differences, and potential missing information that could mitigate disparities. Stereotyping, or ascribing common traits to all individuals in a group, is a particular risk for machine learning models, which generate predictions based on feature similarities. Here, I explore stereotyping in regression or classification models. I demonstrate that common fairness metrics are unable to distinguish stereotyping from decisions based on arguably reasonable factors. Therefore, Paychex has developed a “due diligence” script to guide data scientists in assessing disparities. This process uses model outputs and metrics to identify areas of risk, then Shapley explainers with custom references identify features driving differences. Additional reasoning is suggested to investigate the underlying causes of inequalities, consider the effects of missing information that can mitigate risk, and the effects of sensitive features. This structured report format is suitable for stakeholders or oversight committees.
For most data science applications, the algorithms do not matter nearly as much collecting the right data and preparing that data properly. But sometimes, we really need to pull out all the stops, meaning we need to use complex algorithms to solve difficult problems because of the nonlinearity of the data and the existence of unknown but extensive interaction effects between input variables.This talk will describe situations describe what the simpler algorithm do and what they cannot do, and then provide theory and examples for how complex algorithms, like Random Forests, XGBoost and Deep Learning networks, solve those difficult problems. The talk will also include suggestions for how to identify when your problem may benefit from these algorithms, yet these algorithms, in spite of their strengths, don’t necessarily solve the problem.
Low signal to noise ratios often lead to the inclusion of a large number of variables in Machine Learning models relying on regularization to overcome variance and overfitting problems. Deep learning models also tend to lead to a large number of estimation parameters, but depending on the setup of the problem, the available, or relevant, data points for training may be limited or too costly to obtain. We will discuss and show examples on how we have applied different techniques from feature elimination using backward greedy or feature importance based on SHAP values, to dimensionality reduction using (beta-) Variational Auto Encoders to address the issue.
Anyone can make a pretty bar graph, but can you make sound decisions based on that graph? Is it actionable, or is it only fluff? How do you turn flashy concepts into actionable visualizations? Can you see the end result of those concepts; will they ever become reality? Do you have the vision to combine beauty with brains, thereby driving decisions with data? Or do you settle for destroying direction with disaster? American mathematician John Tukey once said, "The greatest value of a picture is when it forces us to notice what we never expected to see." What value do you see in your data? And what ideas do you have when you see it? Learn how you can capitalize on your ideas by blending internal with external, leveraging them into a cohesive strategy for both the short term AND the long term. See the five "Stages of the Spectrum" in action while discovering the difference between impact and influence, and how that difference plays into making data actionable. Catch the right blend of art and science, or beauty and brains, as you go from concept to reality.
Unplanned downtime is one of the largest sources of lost production in the manufacturing world, leading directly to a loss of possible revenue. The rise of the Internet of Things has led to the development of advanced predictive maintenance solutions aiming to help manufacturers achieve the ultimate goal: zero unplanned downtime. While advances in technology have led to broader adoption of these IoT solutions, many of them are still not accurate enough to rely on. Markus Larsson from Xerox’s Palo Alto Research Center (PARC) talks about what innovators can do to bring IoT predictive maintenance to the next level.
"Can I drive more efficiently?" was the question to be answered at the beginning of all data science efforts at Tracks. How can we accurately measure efficient driving of the thousands of truck drivers on the roads? This is a typical real-life question that is difficult to solve with ML methods. The answer to this question has sparked a number of follow up business questions that we are tackling with Tracks' complex AI system. In this session you will learn how Tracks' solution looks like, which ML methods are employed and which business questions are being answered. In a nutshell, how to turn ML into saved CO2.
Terms of service (ToS) violators at Stripe are merchants on our ecosystem who are selling items and services that are prohibited by our terms of service. This talk will present how we build a multimodal ToS violation deep learning detection system that combines text, images, and tabular data. We will also discuss how we enable interpretability of model predictions.
Image classification has been solved successfully for many tasks e.g. using deep learning techniques. However, in many application scenarios, the set of classes from which input images are drawn is not completely known at time of modeling. It is then important to a) reduce false positives during inference and b) enable the description of "unknown" image content during annotation. This case study shows how we addressed these issues at Miele for recognition of food items together with a concept for dealing with unkown classes during data annotation.
A recurring theme among analytics and data science leaders is the concern of not being able to keep up with all of the rapid change taking place – both individually and as a team. In years past, it was possible to stand up an organization largely made up of analytics generalists who would handle initiatives end to end. In today’s world, it is necessary to utilize a range of specialists focused on either specific methods or specific points in the lifecycle of an analytics and data science initiative. This talk will discuss the trends driving the need to evolve your organization’s talent model, new roles you need to consider implementing, and how they all fit together.
Ashish has distilled his experience of building multiple recommendation system into a set of patterns that he's seen emerge again and again. With a knowledge of these patterns, a practitioner can accelerate their time to successful deployment of a sophisticated recommendation system.
Thomson Reuters produces timely financial news and alerts that is used by many of our corporate and institutional customers. This is extracted from multiple sources, including corporate disclosures, and presented as an ongoing series of alerts and bulletins. One of the major challenges in this kind of financial work is the lack of accurate or weak training data which required us to employ various innovative data cleaning techniques. We also discuss our NLP-based production system that makes strong use of a BERT pre-trained model in combination with an unsupervised strategy, as well as some of the engineering challenges around making the system operate adequately in real-time.
Hospitals make lots of efforts to improve their capacity in emergency departments by adding boxes and beds, organizing shifts, processes and protocols. And, by also implementing systems to record, control and visualize better what's the situation in real time. But decisions, the ones that ultimately drive the output of the process, are entirely left to human capacity. What if an algorithm could help selecting the sequence for attending emergency patients so that life threatening situations are prioritized and overall queueing time is shortened, with no change in the physical resources? Benjamin Arias Gálvez shares the experience at one of the largest public hospitals in Chile, where an algorithm helps sequencing the waiting queue of patients at in the emergency room in real time, with no investment in the physical resources.
Elder Research, a data science consultancy, and Sira-Kvina kraftselkap, a large Norwegian power producer, have deployed deep learning models as part of a predictive maintenance solution to identify likely component failures within hydroelectric generator and turbine systems. In this session, we discuss the design concept, modeling approach, solution architecture, and how it is implemented as part of the maintenance team workflow. We will also explain what makes machine learning so challenging in the utlities industry where labeled cases are in short supply.
As demand for carbon credits accelerates, there is an immense challenge in scaling the supply of carbon offsets. It’s hard to create credits that are additional, non-leaky, and durable, and it’s impossible for all but the largest landowners to participate in carbon programs. Over the last 10 years, SilviaTerra has built technology that generates comprehensive forest inventories of unprecedented resolution and scale, enabling measurement and payment for a comprehensive set of beneficial outcomes across the landscape. This new market is making carbon and other types of natural capital work for all landowners - for every acre, every value, every year.
As online fraud increases in volume, malicious actors rely on automation to keep scaling. To spread false information, sell nonexistent products, or spend money from stolen credit cards, fraudsters use scripts and other automation tools to manage a large number of fake accounts. Such automation introduces a common thread in the profiles or communications of the accounts they control. We developed a deep learning model to detect suspicious patterns amongst different accounts. The model was successful in detecting coordinated attacks, even when deployed on a novel platform, increasing by up to 80% the detection of malicious actors.
Deep learning solutions are nowadays a standard tool in many technological fields. Specifically, in the microbiological field, this is possible through Full Laboratory Automations.
The combination of those two game-changers, made available by COPAN, allow microbiologists to streamline their daily routine: the preparation, incubation and evaluation of thousands of samples (mainly from negative analysis) focusing their high skilled qualities directly on the most challenging and critical ones.
In this talk, Giovanni Turra, key member of the Imaging and Data Analysis team at COPAN, explains how deep learning supports and empowers the daily laboratory battle against diseases.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
.
Machine Learning Week - Virtual - Day 3 - Wednesday, May 26th, 2021
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Models generalize best when their complexity matches the problem. To avoid overfit, practitioners usually trade off accuracy with complexity, measured by the count of parameters. But this is surprisingly flawed. For example, a parameter is equivalent to one "degree of freedom" only for regression; it can be > 4 for decision trees, and < 1 for neural networks. Worse, a major source of complexity -- over-search — remains hidden. The vast exploration of potential model structures leaves no trace on the final (perhaps simple-looking) model, but has outsized influence over whether it is trustworthy.
I’ll show how Generalized Degrees of Freedom (GDF, by Ye) can be used to measure the full complexity of algorithmic modeling. This allows one to fairly compare very different models and be more confident about out-of-sample accuracy. GDF also makes clear how seemingly complex ensemble models avoid overfit, and lastly, reveals a new type of outlier -- cases having high model influence.
Models generalize best when their complexity matches the problem. To avoid overfit, practitioners usually trade off accuracy with complexity, measured by the count of parameters. But this is surprisingly flawed. For example, a parameter is equivalent to one "degree of freedom" only for regression; it can be > 4 for decision trees, and < 1 for neural networks. Worse, a major source of complexity -- over-search — remains hidden. The vast exploration of potential model structures leaves no trace on the final (perhaps simple-looking) model, but has outsized influence over whether it is trustworthy.
I’ll show how Generalized Degrees of Freedom (GDF, by Ye) can be used to measure the full complexity of algorithmic modeling. This allows one to fairly compare very different models and be more confident about out-of-sample accuracy. GDF also makes clear how seemingly complex ensemble models avoid overfit, and lastly, reveals a new type of outlier -- cases having high model influence.
Models generalize best when their complexity matches the problem. To avoid overfit, practitioners usually trade off accuracy with complexity, measured by the count of parameters. But this is surprisingly flawed. For example, a parameter is equivalent to one "degree of freedom" only for regression; it can be > 4 for decision trees, and < 1 for neural networks. Worse, a major source of complexity -- over-search — remains hidden. The vast exploration of potential model structures leaves no trace on the final (perhaps simple-looking) model, but has outsized influence over whether it is trustworthy.
I’ll show how Generalized Degrees of Freedom (GDF, by Ye) can be used to measure the full complexity of algorithmic modeling. This allows one to fairly compare very different models and be more confident about out-of-sample accuracy. GDF also makes clear how seemingly complex ensemble models avoid overfit, and lastly, reveals a new type of outlier -- cases having high model influence.
Models generalize best when their complexity matches the problem. To avoid overfit, practitioners usually trade off accuracy with complexity, measured by the count of parameters. But this is surprisingly flawed. For example, a parameter is equivalent to one "degree of freedom" only for regression; it can be > 4 for decision trees, and < 1 for neural networks. Worse, a major source of complexity -- over-search — remains hidden. The vast exploration of potential model structures leaves no trace on the final (perhaps simple-looking) model, but has outsized influence over whether it is trustworthy.
I’ll show how Generalized Degrees of Freedom (GDF, by Ye) can be used to measure the full complexity of algorithmic modeling. This allows one to fairly compare very different models and be more confident about out-of-sample accuracy. GDF also makes clear how seemingly complex ensemble models avoid overfit, and lastly, reveals a new type of outlier -- cases having high model influence.
Models generalize best when their complexity matches the problem. To avoid overfit, practitioners usually trade off accuracy with complexity, measured by the count of parameters. But this is surprisingly flawed. For example, a parameter is equivalent to one "degree of freedom" only for regression; it can be > 4 for decision trees, and < 1 for neural networks. Worse, a major source of complexity -- over-search — remains hidden. The vast exploration of potential model structures leaves no trace on the final (perhaps simple-looking) model, but has outsized influence over whether it is trustworthy.
I’ll show how Generalized Degrees of Freedom (GDF, by Ye) can be used to measure the full complexity of algorithmic modeling. This allows one to fairly compare very different models and be more confident about out-of-sample accuracy. GDF also makes clear how seemingly complex ensemble models avoid overfit, and lastly, reveals a new type of outlier -- cases having high model influence.
Models generalize best when their complexity matches the problem. To avoid overfit, practitioners usually trade off accuracy with complexity, measured by the count of parameters. But this is surprisingly flawed. For example, a parameter is equivalent to one "degree of freedom" only for regression; it can be > 4 for decision trees, and < 1 for neural networks. Worse, a major source of complexity -- over-search — remains hidden. The vast exploration of potential model structures leaves no trace on the final (perhaps simple-looking) model, but has outsized influence over whether it is trustworthy.
I’ll show how Generalized Degrees of Freedom (GDF, by Ye) can be used to measure the full complexity of algorithmic modeling. This allows one to fairly compare very different models and be more confident about out-of-sample accuracy. GDF also makes clear how seemingly complex ensemble models avoid overfit, and lastly, reveals a new type of outlier -- cases having high model influence.
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Case study of call center analytics work focusing on first call resolution and call reduction. This is a journey through structuring governance around call related projects and value driven approach of prioritizing opportunities as well as solutions. From meta data analysis to NLP models and back once again.
With the need to cater to a global audience, there is a growing demand for applications to support speech identification/translation/transliteration from one language to another. This talk starts off with a brief introduction to the topic of Machine Translation (MT), the evolution/application of Machine Translation. The focus will be on Neural Machine Translation (NMT). Then it moves on to introduce some of the typical customer cases and finally ends with how to embed such functionality in your application. The talk will end with a Python based solution demo and introduce the audience to some resources and tools that could help them as they continue to explore the domain.
The shock of a pandemic can be mitigated by developing a plan for being proactive with modeling that can inform decision making. At Paychex we developed a comprehensive COVID-19 insights package released in March 13 that was then transformed into a real time indicator site that included predictive modeling for client losses.
The health care industry is changing rapidly, therefore, it’s necessary to improve efficiency to production with appropriate and targeted automation. CDPHP, a mid-size payer in New York’s Capital Region, is implementing an Analytics Factory to achieve this end. It does this using a CI/CD/CT framework. In this session, you will learn about our strategy, as well as practical lessons learned on scaling and operationalizing data products. Three use cases will be discussed to illustrate the streamlined process to production: (1) NLP-driven quality measure detection, (2) prioritizing member voice responses for action, and (3) deploying and hosting a readmission model.
With the promise around analytics and AI, many organizational leaders have easily bought into their adoption to drive business growth and impact. For success to be achieved, these leaders need to be committed to the process, creating an organizational foundation to support the development of analytics and AI tied to their business strategy. Commitment (vs buy-in) is much harder to achieve. This talk will explore the differences between organizational commitment and buy-in, what struggles to look out for, and how to develop the commitment to ensure organizational success with data analytic adoption. Sarah will share her experiences of leading and raising organization's analytic maturity, and how the buy-in and commitment differences play into this success.
Climate change is increasing the frequency and severity of natural disasters. Natural catastrophes impact all critical infrastructures, and their resilience is essential for businesses and cities to operate effectively and safely. At One Concern, we combine machine learning and hazard modeling along with ML operational tools to better model the impacts of natural disasters on these critical infrastructures. By taking advantage of modeling, we can understand these potential impacts sooner to plan for and mitigate them. This helps to make our communities more resilient. This session will cover how One Concern applies Machine Learning algorithms to Natural Disaster Modeling.
Graph embedding on very large graphs is a challenging task with many real-world applications such as fraud detection and cross-sell recommendation. The huge computational workload can only be efficiently handled by scalable distributed systems such as Spark or GPU-based deep learning frameworks such as PyTorch. Fugue is an open source Python framework which can manage various backend computing engines with constant interfaces. We will demonstrate how we scale up the transductive embedding algorithm “node2vec” using Spark to handle giant graphs in the Fugue framework. We also applied Fugue and Spark to the inductive embedding algorithm “GraphSAGE” for faster edge sampling.
Carbon Re is a micro AI and Climate Tech startup developing solutions to help Foundation Industry manufacturers, such as cement, steel, chemicals and glass, transition to net-zero. These industries are vital to the global economy, producing 75% of all the material for manufacturing and construction sectors however, today, they represent 21% of global GHG emissions. Carbon Re's first product, the Foundation Platform, is a software platform based on state-of-the-art process improvement techniques for foundation industries developed at the Institute for Manufacturing at Cambridge University (by our co-founder Daniel Summerbell) and built on cutting edge deep reinforcement learning expertise (by our co-founder Aidan O’Sullivan at UCL). Carbon Re's solution means that manufacturers don't have to choose between profitability and sustainability: they can cut their emissions today and improve their finances. This session will cover how Carbon Re combines Machine Learning and Process Improvement techniques to exploit the efficiency opportunities in manufacturing - which is the main path to decarbonisation in the short term.
Every year in Australia ~200 people lose their lives and 100,000+ are seriously injured in potentially avoidable work place incidents. The challenges presented by the extreme conditions and complex work environments typical in heavy industries need to be overcome. This talk presents an advanced computer vision edge AI/ML for these harsh environments. Details are presented on the journey from the R&D group of one of the world’s larger construction companies to a venture capital funded startup; along with the requisite non-trivial performance, market adoptability and technology scalability accomplishments, and the underpinning approaches for Computer Vision & Edge AI/ML.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
In this talk Dr. Anzelc will provide an overview of her own career path from particle physics PhD to insurance analytics executive to people analytics leader, providing a view of how she approached her job search and developed her career path from individual contributor to team leader. Practical advice on what to do, resources available, and other non-academic career paths of physicists will also be shared
Companies are frequently faced with large amounts of unstructured text data, like forum comments or product reviews. Important trends can emerge in these datasets, but it can be time-consuming to read through comments, and keyword matching frequently misses critical nuances. We'll discuss how we've approached this problem at Google using Natural Language Processing, with examples of the approach applied to open datasets. We'll explore how this fits into the ML project lifecycle, with examples of common pitfalls. Finally, we'll highlight how to use this technology as part of a "human in the loop" approach to supercharge your existing team members.
Algorithmic trading revolves around optimal scheduling, when and how to trade each asset of a portfolio. The schedules produced, however, should not be static; being able to optimally adapt to changing intraday market conditions and analysing effectively other exogenous information is of paramount importance. In this talk, we focus on how machine learning methods like RNNs, LSTMs and decision trees, used in conjunction with classical dynamical system techniques, such as kalman filters, can enrich the toolbox available for real-time algorithmic trading.
The speaker will review case studies from real-world projects that built AI systems using Natural Language Processing (NLP) in healthcare. These case studies cover projects that deployed automated patient risk prediction, automated diagnosis, clinical guidelines, and revenue cycle optimization. He will also cover why and how NLP was used, what deep learning models and libraries were used, and what was achieved. Key takeaways for attendees will include important considerations for NLP projects including how to build domain-specific healthcare models and using NLP as part of larger and scalable machine learning and deep learning pipelines in distributed environment.
Machine teaching is a complementary approach to machine learning. It helps those without AI expertise break a complex problem into simpler tasks and give the AI model important clues about how to find a solution quickly. In this session, we talk about the current limitations with control systems and how AI is bridging that gap by training intelligent controllers .
About 30-40% of food produced worldwide is wasted. This represents a $165B loss to the US economy and poses major environmental problems: it is estimated that food waste contributes to up to 25% of all greenhouse gas emissions. This session explores how artificial intelligence can be used to automate decisions across the food supply chain in order to reduce waste and increase the quality and affordability of food. We focus our attention on supermarkets — combined with downstream consumer waste, these contribute to 40% of total US food losses — and we describe an intelligent decision support system for supermarket operators that optimizes purchasing decisions and minimizes losses. The core of our system is a model-based reinforcement learning engine for perishable inventory management. Our system is currently deployed across 220 supermarkets in the US (handling ~2% of US produce volume) and has led to waste reductions of up to 50%. We hope that this talk will bring the food waste problem to the attention of the machine learning community.
Online fraud is flourishing as online services extend to more industries, including financial service providers, insurance companies, online retailers, and social networks. Network traffic data is invaluable for identifying such malicious actors. In this talk, we will show how to train a self-supervised deep learning model using only unlabeled network logs. Deriving the representation from the "structure" of the bits in an IP address, the model can process novel entities not encountered in the training data. Results show that our proposed framework can identify anomalous network accesses up to 16 times better than Amazon SageMaker's state-of-the-art model.
Since 2018, Transformer models revolutionized the planet NLP by bridging the gap between universal and task-specific representations. Transformers accelerated the shift from a world in which each NLP application requires training deep neural networks from scratch to the emergence of large pretrained models with great generalization capabilities, which are finetuned for downstream tasks with limited amount of labelled data. Yet, using Transformers for industrial applications poses various challenges. In this accessible presentation, Hadrien Van Lierde from Tencent-backed WeBank will describe how China’s leading digital bank uses Transformers to improve its customer service chatbot while maintaining low latency and high throughput.
Machine Learning Week - Virtual - Day 4 - Thursday, May 27th, 2021
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Grab your real coffee and share experiences virtually with your peers to explore the new challenges of operating in a largely virtual world. Just like pre-show breakfast in a regular conference you’ll join a “round table” with seven fellow attendees and see where the conversation takes you.
Despite some dystopian predictions, Machine Learning and AI are not about to replace humans completely. But there's concern and confusion about how best to integrate people and machine learning algorithms. Should you have your algorithms recommend actions and allow people to override them? Should you balance human judgment alongside your algorithms, integrating them into a coherent whole? What about capturing the know-how and experience of your people so you can embed it in the algorithm itself? There are many ways to proceed, each with trade-offs around accuracy, organizational adoption, ethics and legality.This panel will discuss the various ways you can combine people and algorithms and maximize the value of both in your operations.
Despite some dystopian predictions, Machine Learning and AI are not about to replace humans completely. But there's concern and confusion about how best to integrate people and machine learning algorithms. Should you have your algorithms recommend actions and allow people to override them? Should you balance human judgment alongside your algorithms, integrating them into a coherent whole? What about capturing the know-how and experience of your people so you can embed it in the algorithm itself? There are many ways to proceed, each with trade-offs around accuracy, organizational adoption, ethics and legality. This panel will discuss the various ways you can combine people and algorithms and maximize the value of both in your operations.
Analytics adoption in healthcare has come a long way over the last two decades. In the early years, advanced machine learning was limited to a handful of experts working on narrowly-defined subjects such as hospital readmissions. Among most healthcare professionals, there was limited understanding of what machine learning is, and even less eagerness to integrate analytic models into medical practice or organizational operations in any meaningful way. Today, any healthcare organization that is not using advanced analytics is an anomaly. Analytic methods are being employed almost everywhere in healthcare from insurance to precision medicine with tailored treatment plans guided by artificial intelligence. And, analytics are embedded in the systems used by hospitals, such as the HIS or the laboratory technology. Slowly, advanced analytics is being integrated throughout the healthcare system even though many professionals are not aware of its influence. Join our expert panel as we discuss the current state of analytics/machine learning, the work that remains, and opportunities to better use this incredible technology. Attendees are invited to join in the conversation and contribute questions and comments to the discussion.
As organizations increase their maturity in the adoption of machine learning and AI, they are learning important lessons along the way. In this keynote presentation, Steven Ramirez, CEO of machine learning consultancy Beyond the Arc, will focus on 3 key takeaways:
6 steps to accelerate the pace of AI adoption
Ways to level-up your capabilities with NLP, open stack tools, and cloud-native approaches
Cultivating talent for AI progress
Discover how early-stage climate tech companies are using machine learning to help meet their challenges.
Deep learning and Computer Vision are changing the way to improve product quality in manufacturing industry. We have used cutting edge neural network architectures to identify the source of problems in products. It improves the product quality by improved defect detection, defect categorization and enhances the customer experience. The models are deployed into production and are generating fantastic results. It will be surely a great attraction for the visitors who want to analyze the significance of deep learning, identify the process and challenges. It will be a first hand information for them which will prove to be really useful and will have far-reaching results.
In this presentation, Mark Do Couto, Senior Vice President Data Analytics at Altair, discusses the complexities of Machine Learning and AI. Join this session to hear how Altair is bringing Data Analytics to the masses through a No Coding environment. The future for ML/AI is now and Altair is here to show you how everyone can take advantage of it.
Deep learning models for forecasting and planning have shown significant promise for handling multiple variables, uncovering hidden patterns, and producing accurate forecasts. However, as one might expect, deep learning models are also complex and rife with pitfalls. Since these techniques often seem like a ‘black box,’ managers -- both technical and nontechnical backgrounds -- can find them hard to master.
In this session, Senior Data Scientist, Javed Ahmed will focus on the intuition behind various deep learning approaches, explore how managers can tackle highly complex models by asking the right questions, and evaluating the models with familiar tools.
Attendees at the Metis session will leave with the tools to:
● Identify types of forecasting applications that can benefit from deep learning
● Broadly understand deep learning approaches relevant to forecasting
● Understand pitfalls related to deep learning approaches, and why simpler models may work better
● Evaluate the results of a forecasting program
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
About 80% of machine learning projects fail – not because the data scientists failed, but because of the many known and unknown unknowns especially of data, but also of business and users. Consequently, the key to success is to “fail early to succeed sooner” i.e., to identify the critical break points and to focus on the use case with high chances of success. The identification of the right use cases is the purpose of a data strategy and data thinking is the process of thinking through the right solution. This interactive session introduces a method for data thinking, called Data Strategy Design. The free Data Strategy Designkit consists of 15 visual tools to collaboratively and interdisciplinary develop data strategies and data products. It is already used by hundreds of companies worldwide and by participating in this deep dive session you will also learn how to apply the method and tools to your projects. For the joint exercise we will use Miro, an online whiteboard platform, and we will work together on an exemplary machine learning project.
While hotels' star ratings are widely recognized and used across the globe, there are no similar rating system exists for the large majority of vacation rentals, making it difficult for guests to search and compare options and hard for vacation rental suppliers to market their product effectively.
In this talk, we will present a Quality Rating System for Vacation Rentals based on Machine Learning, focusing on automated explainable vacation rental quality ratings. This system was successfully deployed and validated through Randomized Controlled Trials at Booking.com, impacting millions of accommodations and guests.
The Safety National Casualty Corporation is the leader in Excess Workers' Compensation. Predictive Analytics is used internally to identify those claims that have potential of exceeding high-cost levels in years. Structured claim data, such as claimant age, nature of injury and body parts, provide fundamental features for predicting claim severity. However, in many cases, they are not informative and distinguishing enough to identify extremely sever claims which are rare in the claim data pool. To address this issue, we explore the unstructured claim data – claim notes, and investigate how different text mining techniques can be used to extract predictive insights from claim notes.
OSF Healthcare has been successfully internally developing and deploying advanced analytics solutions for the past eight years. During that time, our work has been tightly focused on a limited set of core clinical areas of our service Ministry. Driven in large part by the success achieved through this work, OSF executive leadership made the strategic decision to expand Advanced Analytics offerings to become truly enterprise scale. This session will showcase the structure and service offering adjustments we’ve made over the past year to support this transformation. The speaker will also highlight specific projects empowered through these adjustment.
Building a strong AI/Machine Learning Community means more than just checking off a box. It means enabling your workforce with critical technical skills, promoting the adoption of advanced analytics, and providing overall faster customer response times. It means creating an inclusive space where your employees feel heard and are able to share best practices, current issues, and new project ideas and technologies.
Dive deeper with us into what it takes to get a community set up, where some common pitfalls occur, the expected benefits for stakeholders, and finally some tips and tricks for growing your community and helping it thrive!
Methane, the primary component of natural gas, is responsible for 15% of global warming. Our mission of finding and stopping greenhouse gas emissions at huge scale is a critical step towards controlling climate change, but it also presents unique challenges. And, as a small startup, navigating the trade-offs between speed, accuracy, and cost in our data pipeline can often be the difference between survival and failure. In this talk, we will examine the difficulties ML pipeline design in cases where information, time, and money are constrained, and how to do so while hiding the sausage-making from our customers, who just want to know where their equipment is leaking, and want to know fast. By using a lean, iterative approach that involves input from every department, including engineering, operations, and business development, we stay focused on creating analytics that maximize value while reducing risk to the company.
For machine learning problems, we don't lack great tools for certain problems, we lack a unified approach to use them for both prototyping and production. Fugue is a framework, aiming to bridge this gap. In this talk, we are going through a real deep learning example with preprocessing, training, and hyperparameter tuning. We will discuss the pain points and demonstrate how you can use Fugue to quickly iterate on small data and scale up on a Spark cluster without code change. We may also talk about some Lyft use cases, and how Fugue changed the game.
Take a break or join your fellow practitioners in an interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Modeling the risk of customer churn in large B2B organizations can be tricky. In this session, we will outline our approach to handling that complexity by developing an interpretable model and an interactive front-end tool to highlight risk factors to a team of non-technical stakeholders. We highlight the importance of transparent modeling (as opposed to black-box modeling) in B2B retention, where understanding the risk factors is as important to our customer experience stakeholders as the model itself.
Systemic racism is a complex and multi-dimensional problem that manifests in several different ways. No single narrative or data point can fully capture the pervasive and damaging nature of this crisis. Through our work at Cape Fear Collective, we attempt to decompose the far-reaching impacts of systemic racism into tangible components. The culmination of this work is a Racial Equity Index, which we believe will aid in the development of strategies and solutions to combat systemic racism and forge a more equitable future.
Predictive Analytics is transforming commercial insurance underwriting. It delivers more accurate and real-time risk assessment which helps underwriters efficiently acquire information, prepare more accurate quotes, and speed up and even automate the underwriting process. This talk presents one practical application of predictive analytics on building the triage process to help underwriters prioritize their attention on the most valuable submissions.
Counterfeited, adulterated, and stolen pharmaceuticals are threats to US citizens. The Drug supply Chain Security Act (DSCSA) was established to secure the supply chain through various means including the establishment of an interoperable traceability system. The current industry stakeholder design is a distributed structure with exchanges between established trading partners. While this is flexible, it creates challenges to detecting nefarious activity and performing supply-demand management. We have implemented an interoperable, traceable prototype system which solves some of the problems and fulfills DSCSA requirements that must be met by 2023 .
Enhanced accessibility and availability to data has resulted in the increasing demand for data scientists within the logistical and supply end of industry. Yet, with this great reservoir of information, data scientists are faced with key challenges in the development of their solutions. In this session, we explore both the traditional as well as new challenges of the data scientist in this environment. The alignment of the right data infrastructure in solving the right business problem is just one example. New data sources and new tools exacerbate this issue but the disciplined process approach towards the data science process is the constant. By looking at a variety of industries such as insurance, travel , and health, we examine how to develop the best practices within this process. A number of examples and case studies will be presented to explore the increased shift and emphasis towards data science skills that focus less on the technical and more on identifying the business problem.
Globally, more than $200 billion worth of recyclable materials goes unrecovered annually. The economics and efficiency of identifying and sorting paper, plastics, metals, and other recyclables from the waste stream creates a major challenge for material recovery. In recent years, the waste industry has also faced stricter international quality standards for contamination-free imports of recycled materials, leaving the industry in search of cost-effective alternatives to meet these requirements. COVID-19 then forced many businesses to suspend recycling operations due to concerns for worker safety. Simultaneously, the pandemic increased demand for high-quality recycled feedstock to overcome supply chain interruptions and shifts in raw material availability.
Learn how AMP Robotics’ technology, which applies computer vision and deep learning to identify and differentiate recyclables found in complex, mixed material streams, is helping the waste industry meet these challenges by modernizing recycling—improving material quality, ensuring worker safety, increasing productivity, lowering costs, diverting waste from landfill, and reducing greenhouse gas emissions—while increasing overall rates of recycling and resource recovery.
Many organizations deal with incoming customer problems either internally or externally. Frequently, these problems are handled one at a time by humans. We’ll discuss how at Google we are doing early detection of broad IT incidents before they become too large. We'll go through unsupervised learning techniques that when combined with deep learning-based language modeling creates a powerful, robust system that has saved Googlers hundreds of thousands of hours in productivity.
Thomson Reuters produces timely financial news and alerts that is used by many of our corporate and institutional customers. This is extracted from multiple sources, including corporate disclosures, and presented as an ongoing series of alerts and bulletins. One of the major challenges in this kind of financial work is the lack of accurate or weak training data which required us to employ various innovative data cleaning techniques. We also discuss our NLP-based production system that makes strong use of a BERT pre-trained model in combination with an unsupervised strategy, as well as some of the engineering challenges around making the system operate adequately in real-time.
Machine Learning Week - Virtual - Day 5 - Friday, May 28th, 2021
A global NGO identifies and rehabilitates tens of thousands of survivors of human trafficking and slavery situations. Different situations necessitate different rehabilitation “journeys.” We analyzed case worker notes from check-ins with survivors to help understand how specific programs, and especially “trauma-informed” programs, might be affecting survivors’ emotional wellbeing and likelihood to finish their journeys. We found generally negative associations between survivor sentiment and interactions with the justice system, and generally positive associations with training programs, especially for rebuilding social support networks and economic empowerment. These findings have implications for journey planning as the NGO tries to scale up its worldwide operations.
A global NGO identifies and rehabilitates tens of thousands of survivors of human trafficking and slavery situations. Different situations necessitate different rehabilitation “journeys.” We analyzed case worker notes from check-ins with survivors to help understand how specific programs, and especially “trauma-informed” programs, might be affecting survivors’ emotional wellbeing and likelihood to finish their journeys. We found generally negative associations between survivor sentiment and interactions with the justice system, and generally positive associations with training programs, especially for rebuilding social support networks and economic empowerment. These findings have implications for journey planning as the NGO tries to scale up its worldwide operations.
Data scientists and management have different ways of thinking and for good reason: their jobs are quite different! Moreover, data scientists, for all of their strengths, are often not the best communicators to business leaders. These differences, unfortunately, can interfere with the success of analytics projects. A particular problem for non-technical management is understanding how to help data scientists to focus on solving the business problem in the right way.
This talk will raise five questions that should be asked of data scientists. These questions are critical for setting up the problem properly and assessing the models in a manner commensurate with the business objectives.
Reducing unplanned downtime requires the proactive maintenance of assets to keep operation running smoothly. Unfortunately, many organizations rely on time-based or usage-based maintenance rather than actual need to determine when to best schedule a maintenance activity. Predictive maintenance instead uses data from operations to determine the health of the asset and can be used to determine when it is best to perform maintenance. We developed intelligent digital twin technology that utilizes sensor data to assess the health and prognostics of a system based on predictive models built with physics-based simulation data. A digital twin is a digital representation of a physical asset that acts as a “single source of truth” for the asset. By incorporating predictive analytics into a digital twin, we can calculate maintenance-related parameters that would allow for appropriate intervention. In this session we will discuss some of the technology available to implement digital twins. We will also discuss how to augment the basic capabilities of digital twins with predictive algorithms for condition-based maintenance. In particular we will discuss developing a digital twin of a naval vessel that assesses the efficiency of various propulsion system components to determine the state of decay of the components.
Join our panel to find out how established industrial companies are using machine learning to address climate challenges.
Deep Learning image classifiers represent a breakthrough in image recognition and classification tasks. However they come with their own quirks and kinks: A "clean" product image can be easily classified, but what about a stock image of "a woman running in the rain"? Is it a raincoat? Her shoes? The phone cover? At Shopify we were tasked with mapping a large, complex, and "dirty" catalogue of products into duplicates, similar products and categories. In this case study I will walk you through our journey and how we were able to harness the strengths of CNNs while avoiding the major pitfalls.
There are many non-profit organizations trying to help develop and build economic opportunities for families in developing countries around the world by aiding individuals in starting and maintaining small businesses. Vision Fund International is one such organization that uses microfinance loans to jump start business opportunities in these countries. However, these developing countries do not have the credit institutions that countries in the developed world have, which leaves banks restricted on their ability to make loan decisions. Analytical modeling approaches are widely used in major banks in developed countries to efficiently make these decisions. This talk details a project worked on in partnership with Vision Fund International that develops scorecard models for use in loan decisions and the impact it is having on the growth of economies around the world.
There are many non-profit organizations trying to help develop and build economic opportunities for families in developing countries around the world by aiding individuals in starting and maintaining small businesses. Vision Fund International is one such organization that uses micro-finance loans to jump start business opportunities in these countries. However, these developing countries do not have the credit institutions that countries in the developed world have, which leaves banks restricted on their ability to make loan decisions. Analytical modeling approaches are widely used in major banks in developed countries to efficiently make these decisions. This talk details a project worked on in partnership with Vision Fund International that develops scorecard models for use in loan decisions and the impact it is having on the growth of economies around the world.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Most business leaders claim to be "results driven," an orientation reflected in their embrace of a "pay for performance" philosophy. Many performance measures commonly used are “noisy” or contaminated by situational factors far removed from employees’ control. Hence, they fail to capture the actual value that employees deliver, with unintended, negative consequences for employer and employees.
Using case examples, the presenter will show how advanced analytics can address this problem by adjusting performance measures to remove the effects of random or situational factors. By distinguishing employee “value” from “performance.” organizations can deliver more effective incentives at less cost to shareholders.
Webex Contact Center is a multi-tenanted Cloud Contact Center solution from Cisco, that scales to thousands of concurrent interactions. Improving Agent Performance directly impacts the efficiency and effectiveness of Contact Centers. We are building the next generation AI enabled interaction experiences to create super agents. With the help of AI, agents can now focus on the customer, understand their sentiment and answer questions quickly. AI capabilities that were built include Speech-To-Text transcriptions, sentiment analysis, Question-Answering and Topic Modeling. AI Models by themselves don't become a scalable solution. Serving these models at scale, specifically for a multi-tenanted cloud service, takes additional planning, architectures, capabilities and operations. This session discusses the unique challenges, solutions and takeaways in building and deploying these models and solutions. Key topics addressed are:
- Multi-tenanted Cloud Services: unique considerations
- Model Scaling Challenges for real time
- Architecture options for scaling Models
- Adapting Models to serve at scale
- Tuning Models for Resource & Cost scaling
Our Next Best Action journey commenced with the simple aspiration of delivering smarter personalised experiences at the moment of need for our customers. While the aspiration was simple, orchestrating NBA is a long-distance race often fraught with challenges, twists in the road, and steep hills to climb. There are many moving parts to orchestrate the right experiences specially when customers adapt their own non-linear paths, as they move through the different aspects of life events. But how does one set in motion a truly effective execution? In this session, Harphajan would share the art & science of his journey in executing Next Best Action strategy within retail financial services especially where customer relationships evolve over long periods of time
From the beginning of the global pandemic, the efforts to make COVID-19 case and death reports and other related data available to the public have been immense. This raised the hope of ambitious modelers worldwide to understand the disease and what policy measures would be effective in mitigating the pandemic. But immense problems with both the data and the modeling techniques led to false expectations, false claims, and in the end, mistrust of scientific claims generally.I will review some of the misqueues and analytic successes, demonstrating the key differences. We will see how, despite crippling weaknesses inherent in the publicly available data, sound analytical modeling techniques could reliably reveal actionable early insights for the pandemic, and coincidentally, influenza. The analytic lessons learned from this global experience can and must inform health analytics in the future.
Webex Contact Center is a multi-tenanted Cloud Contact Center solution from Cisco, that scales to thousands of concurrent interactions. Improving Agent Performance directly impacts the efficiency and effectiveness of Contact Centers. We are building the next generation AI enabled interaction experiences to create super agents. With the help of AI, agents can now focus on the customer, understand their sentiment and answer questions quickly. AI capabilities that were built include Speech-To-Text transcriptions, sentiment analysis, Question-Answering and Topic Modeling. AI Models by themselves don't become a scalable solution. Serving these models at scale, specifically for a multi-tenanted cloud service, takes additional planning, architectures, capabilities and operations. This session discusses the unique challenges, solutions and takeaways in building and deploying these models and solutions. Key topics addressed are- Multi-tenanted Cloud Services: unique considerations- Model Scaling Challenges for real time- Architecture options for scaling Models- Adapting Models to serve at scale- Tuning Models for Resource & Cost scaling
Financial institutions are playing an increasing role in the low-carbon transition by taking steps to accurately estimate, price, and disclose future climate risk. By quantifying their exposure to climate risks, financial institutions can more effectively allocate investments, avoid ‘stranded’ assets, and track adherence to Paris Agreement goals and shareholder commitments. However, it remains difficult for these institutions to assess climate related risks across a portfolio of assets and across different benchmark warming scenarios.I will cover large scale data transformation approaches as part of an end-to-end framework for quantifying annual, asset-level climate risk over multiple climate hazards including wildfires, inland flooding, and heat waves using simulations from global climate models participating in the Coupled Model Inter-comparison Project Phase 6 (CMIP6).We will be discussing techniques to quantify forward looking climate risk from 2020 to 2050 under multiple climate scenarios such as high-emissions (SSP5-8.5) and medium-emissions (SSP2-4.5) warming scenarios. I will also showcase intermediate steps to make the climate simulations and spatiotemporal data interpretable and actionable. We will cover ways to harmonize near real time observations from ground measurements and satellite derived data with forward looking climate risk projections for acute physical hazards for high accuracy predictive modeling.
In today's world of elastic cloud/on-prem resources to support deployments, it is essential to be right-sized. We present a Hybrid Deep Learning and Statistical approach to model future demand. An accurate measure of incoming demand enables us to be right-sized while keeping guarantees on reliability, resiliency, and availability. We will share our experience dealing with demand volatility and how we temper them to enable actionability with tradeoffs.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Take a break or join your fellow practitioners in this interactive session where you can exchange approaches to shared challenges and hear how your peers are tackling similar issues.
Many analytics teams wait too long to start working on deployment. The most common reason they give is that they want to make sure that the model is “good enough” before they start to plan deployment. The truth is that doing so virtually guarantees project failure, or at a minimum, massive delays. Projects don’t fail because of a failure to find patterns in the data. Projects fail because models that fail to align perfectly with organizational priorities will never be embraced by the end-users of the deployed solution.
In this talk we are going to walk through a design of the typical operating recommender system that operates at a large scale. As part of this talk we are going to do a few deep-dives in the various aspects of the recommender systems and will cover various topics that are critical for the recommender system design, but usually do not receive enough attention, such as:
· Desired properties for multi-stage ranking system
· Sparse features for ranking models
· Limitations of the current approaches used in common models and possible ways to address those
One of the critical trends that is impacting customer experience is driving personalization and relevancy in various customer touch points. Customer expectation is being reset everyday thanks to ultra-personalized experiences driven by Google, Facebook and now even some FinTechs. There is a constant hunt for 'better data' that can enhance our understanding of the customer preferences. Payments, in addition to being a means of closing commerce transactions, is in fact, a very rich source of data. Especially for credit and debit cards (which enjoy extremely high usage in NA), every single transactions comes with rich metadata (point of sale device, use of mobile wallets, merchant categories, location, etc.). This spend data provides unique insights into customer preferences and trends and can be transformed into actionable personas. Merchants and Financial Institutions can leverage this data to build personalized digital customer experiences that can delight the customer translating into high customer value. This talk goes over the structure on how we should look at the digital customer experience stack and how payments data & analytics layer can enhance the customer experience.
Germany is a “super-aged” society with increasing demands placed on the healthcare system due to the rise in chronic diseases. The country increasingly leverages e-health solutions to accommodate a healthier older population. In conjunction with Germany’s largest health insurer Techniker Krankenkasse, we are conducting a pilot study in which we deploy predictive modeling to identify elderly at risk of emergency ambulance transport based on e-health data. A case manager reaches out to predicted high-risk patients and recommends interventions. Initial results demonstrate a significant reduction in ambulance dispatch rate. This presentation will cover predictive model development, deployment and preliminary findings.
In this talk we are going to walk through a design of the typical operating recommender system that operates at a large scale. As part of this talk we are going to do a few deep-dives in the various aspects of the recommender systems and will cover various topics that are critical for the recommender system design, but usually do not receive enough attention, such as:
· Desired properties for multi-stage ranking system
· Sparse features for ranking models
· Limitations of the current approaches used in common models and possible ways to address those
We give an overview of recent developments in physics-informed AI and big data that are transforming the prediction of climate and weather in applications ranging from climate risk modeling for insurance to real-time forecasting for energy. Traditional climate and weather models require computationally expensive simulation of physical laws on supercomputers with hours to days of processing time and have limited capacity to incorporate ground-truth data sources. The development of cloud-based AI workflows based on deep neural networks provides an alternative approach to develop physical emulators of climate and weather processes that are highly scalable and natively tuned to utilize the petabytes of remote-sensing, ground-based and numerical simulation data from Earth observation that are generated daily. We present work that we are doing at Terrafuse AI, a startup out of Berkeley National Lab, to develop an AI-native climate risk and forecasting platform for problems ranging from high-resolution mapping of wildfire risk in California to real-time wind forecasting for aviation and renewable energy.
Combining automated speech recognition with natural language processing allows one to build semi-automatic note taking apps. With a minimal input from the call participants, it would create transcribe speech on the fly and extract and highlight intents and entities, thus creating a semi-structured meeting minutes, ready to be exported to structured databases. For practical applications, it is important for the system to recognize custom lexicon, without expensive acoustic model retraining, solely by automatically building custom language models.
The number of Earth observation satellites has increased significantly and these satellites are producing huge volumes of images at different spatial resolutions and spectral bands. Manually understanding all the pixels is nearly impossible, as it would take years simply to look at them. This talk will cover the challenges and approaches involved in understanding satellite imagery, which are different than those involved in working with other real-world images. At Orbital Insight, we transform multiple sources of geospatial data - including satellite images - into actionable insights to understand economic, societal, and environmental trends at global, regional, and hyper-local scales.
