Log in to bookmark your favorites and sync them to your phone or calendar.

Deep Learning Stage [clear filter]
Thursday, January 24


Compère Welcome - Deep Learning Stage
Our compère for the Deep Learning Stage today will be Anirudh Koul, Head of AI & Research at Aira.

avatar for Anirudh Koul, Aira

Anirudh Koul, Aira

Head of AI & Research, Aira
Anirudh Koul is the Head of AI & Research at Aira (Visual interpreter for the blind), and upcoming author of 'Practical Deep Learning for Cloud and Mobile'. Previously at Microsoft AI & Research, he founded Seeing AI App - often considered the defacto app in the blind and low vision... Read More →

Thursday January 24, 2019 9:00am - 9:15am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Beyond Supervised Driving
Crowd-sourced steering does not sound as appealing as automated driving. We need to go beyond supervised learning for automated driving, including for computer vision problems seeing great progress with strong supervision today. First, we will motivate exciting scientific problems that have huge implications in the research and development of long-term large-scale autonomous robots, such as unsupervised domain adaptation, self-supervised learning, and robustness to edge cases. Second, we will talk about the robotics system perspective, especially end-to-end vs modular design and human-robot interaction. Finally, we will describe some of TRI's related research directions, especially around world-scale cloud robotics. In particular, we will discuss recent related results obtained in the ML team at TRI on state-of-the-art methods for self-supervised depth and pose prediction from monocular imagery, end-to-end panoptic segmentation, and large scale distributed deep learning on GPUs in the cloud.

avatar for Adrien Gaidon, Toyota Research Institute

Adrien Gaidon, Toyota Research Institute

Machine Learning Lead & Snr. Research Scientist, Toyota Research Institute
Adrien Gaidon is the Manager of the Machine Learning team and a Senior Research Scientist at the Toyota Research Institute (TRI) in Los Altos, CA, USA, working on open problems in world-scale learning for autonomous driving. He received his PhD from Microsoft Research - Inria Paris... Read More →

Thursday January 24, 2019 9:15am - 9:35am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Object-oriented Perception and Control
Why are infants better at sensory motor tasks than our current AI systems? We are born with learning mechanisms to map our sensory experience into objects and abstractions over them. My talk will present unsupervised approaches integrating deep/reinforcement learning and probabilistic programming, to learn about objects and goal-directed control grounded in them. I will give demonstrations in a few domains including video understanding, game playing and robotics.

avatar for Tejas Kulkarni, DeepMind

Tejas Kulkarni, DeepMind

Research Scientist, DeepMind
I am a Research Scientist at Google DeepMind. Previously, I was a PhD student at MIT under the supervision of Joshua Tenenbaum. I am primarily interested in understanding how the mind works. My current research goal is to build learning algorithms that acquire grounded common-sense... Read More →

Thursday January 24, 2019 9:35am - 10:00am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Adversarial Machine Learning
Most machine learning algorithms are based on optimization: given a cost function, the algorithm adapts the parameters to reduce the cost. Adversarial machine learning is instead based on game theory: multiple "players" compete to each reduce their own cost, often at the expense of other players. In this talk I show how adversarial machine learning is related to many active machine learning research areas.

avatar for Ian Goodfellow, Google Brain

Ian Goodfellow, Google Brain

Staff Research Scientist, Google Brain
Ian Goodfellow is a Staff Research Scientist at Google Brain. He is the lead author of the MIT Press textbook Deep Learning. In addition to generative models, he also studies security and privacy for machine learning. He has contributed to open source libraries including TensorFlow... Read More →

Thursday January 24, 2019 10:00am - 10:25am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Combining Planning and Learning
In order to build agents that can reason about the consequences of their actions and control the world, it is important to acquire rich priors that capture a notion of how to plan forward into the future. Imagining the consequences in raw pixels is unlikely to scale well for large environments and also has no incentive to ignore aspects of the raw sensory stream that are irrelevant to the task at hand. I will talk about ways to circumvent this issue by introducing explicit differentiable planning inside the policy's computation graph and show that the learned priors are generalizable across different robot morphologies and can capture a generic notion of the underlying task in its representation.

avatar for Aravind Srinivas, UC Berkeley

Aravind Srinivas, UC Berkeley

Ph.D. Student, UC Berkeley
Aravind is a second year Ph.D. student at UC Berkeley advised by Prof. Pieter Abbeel and is part of the Berkeley AI Research lab. He has spent time at OpenAI and is interested in learning representations from raw sensory data for general intelligence.

Thursday January 24, 2019 11:05am - 11:25am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Go-Explore: A New Type of Algorithm for Hard-exploration Problems
A grand challenge in reinforcement learning is producing intelligent exploration, especially when rewards are sparse or deceptive. I will present Go-Explore, a new algorithm for such ‘hard exploration problems.’ Go-Explore dramatically improves the state of the art on benchmark hard-exploration problems, enabling previously unsolvable problems to be solved. I will explain the algorithm and the new research directions it opens up. I will also explain why we believe it will enable progress on previously unsolvable hard-exploration problems in a variety of domains, especially the many that harness a simulator during training (e.g. robotics). More information can be found at https://eng.uber.com/go-explore

avatar for Jeff Clune, Uber AI Labs

Jeff Clune, Uber AI Labs

Senior Research Scientist & Founding Member, Uber AI Labs
Jeff Clune is the Loy and Edith Harris Associate Professor in Computer Science at the University of Wyoming and a Senior Research Scientist and founding member of Uber AI Labs. He focuses on robotics, reinforcement learning, and training neural networks either via deep learning or... Read More →

Thursday January 24, 2019 11:25am - 11:45am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Deep Learning for Robotics and Robotics for Deep Learning
Large-scale, data-driven techniques for robotic learning have imbued robots with unprecedented capabilities in recent years. However, a long-standing challenge for goal-directed learning on robots is the problem of supervision: how do we know if the robot actually achieved some goal, such as picking up the correct object? I present recent work in which we use robotic interaction to provide free supervision signals for deep representation learning, and reuse those very same representations to learn instance grasping. This synergy leads to interpretable visual representation learning and useful grasping skills, freeing us from ever having to label any data.

avatar for Eric Jang, Google Brain Robotics

Eric Jang, Google Brain Robotics

Research Engineer , Google Brain Robotics
Eric is a research engineer on the Google Brain team, working on robotic grasping and manipulation. He is interested in meta-learning for robotics, deep generative models, and Artificial Life. He received his M.Sc. in CS and Bachelors in Math/CS at Brown University in 2016.

Thursday January 24, 2019 11:45am - 12:05pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Recent Advances From OpenAI
avatar for Ilya Sutskever, OpenAI

Ilya Sutskever, OpenAI

Co-Founder & Chief Scientist, OpenAI
Ilya Sutskever received his PhD in 2012 from the University of Toronto working with Geoffrey Hinton. After completing his PhD, he cofounded DNNResearch with Geoffrey Hinton and Alex Krizhevsky which was acquired by Google. He is interested in all aspects of neural networks and their... Read More →

Thursday January 24, 2019 12:05pm - 12:30pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Probabilistic Deep Ensembles for Predictive Uncertainty Estimation
Quantifying predictive uncertainty in deep learning is a challenging and yet unsolved problem. Predictive uncertainty estimates are important to know when to trust the model's predictions, especially in real-word applications, where the train and test distributions can be very different. Bayesian neural networks are currently the state-of-the-art for estimating predictive uncertainty; however these require significant modifications to the training procedure and can be computationally expensive. I'll talk about our recent work on "Probabilistic Deep Ensembles", an alternative to Bayesian neural networks, that is simple to implement, readily parallelizable, requires very little hyperparameter tuning, and yields high quality predictive uncertainty estimates. I'll discuss experiments that show that our method produces well-calibrated uncertainty estimates and is robust to dataset shift, and also highlight how we used this method in a challenging healthcare problem.

avatar for Balaji Lakshminarayanan, DeepMind

Balaji Lakshminarayanan, DeepMind

Senior Research Scientist, DeepMind
Balaji Lakshminarayanan is a senior research scientist at Google DeepMind. He's interested in scalable probabilistic machine learning and its applications. Most recently, his research has focused on probabilistic deep learning, specifically, uncertainty estimation and deep generative... Read More →

Thursday January 24, 2019 1:30pm - 1:55pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Applying ML & NLP in Google Ads
Building and deploying machine-learning (ML) models at Google comes with interesting challenges. For example, some models have to handle massive amounts of training data, while some supervised tasks have insufficient amount of training labels. Or, even when the model quality is good enough for a product requirement, it may not meet other requirements (e.g., serving latency, memory footprint). In this talk we will discuss some of these challenges and share our experiences from deploying ML models for quality improvements in Search Ads products via some case studies. One particular case study I will discuss in detail is a recent paper where we use deep neural networks to understand ad performance and attribute it to particular parts of ad text. This is an interesting research problem in Natural Language Processing (NLP) -- we will outline our key results related to this problem, and discuss interesting areas of future research.

avatar for Sugato Basu, Google

Sugato Basu, Google

Senior Staff Research Scientist/Tech Lead of AdsAI, Google
Dr. Sugato Basu is currently the Tech Lead of the AdsAI team in Google, which applies state-of-the-art machine learning (ML) and natural language processing (NLP) technology to challenging problems in Search Ads at Google. He joined Google in 2007 and has worked for more than a decade... Read More →

Thursday January 24, 2019 1:55pm - 2:20pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


On-device Neural Networks for Natural Language Processing
Deep neural networks reach state-of-the-art performance for wide range of Natural Language Processing, Computer Vision and Speech applications. Yet, one of the biggest challenges is running these complex networks on devices with tiny memory footprint and low computational capacity such as mobile phones, smart watches and Internet of Things. In this talk, I will introduce novel on-device Self-Governing Neural Networks (SGNNs), which learn compact projection vectors with local sensitive hashing. The key advantage of SGNNs over existing work is that they surmount the need for pre-trained word embeddings and complex networks with huge parameters. I will showcase results from extensive evaluations on wide range of natural language tasks such as dialog act classification and user intent prediction. Our findings show that SGNNs are effective at capturing low-dimensional semantic text representations, while maintaining high accuracy and improving over state-of-the-art results.

avatar for Zornitsa Kozareva, Google

Zornitsa Kozareva, Google

Manager, Google
Dr. Zornitsa Kozareva is a Manager at Google, leading and managing the Natural Language Understanding group and efforts in Google Apps Intelligence. Prior to that, Dr. Kozareva was Manager of Amazon’s AWS Deep Learning group that built and launched the Natural Language Processing... Read More →

Thursday January 24, 2019 2:20pm - 2:40pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Continuous Object Detection for Conversational Vision
Object detection is a core computer vision task, where a machine learning (ML) model is trained to identify objects from a pre-specified set of object categories. In a real-life scenario, e.g., when an object detector is used to process the picture taken by a mobile phone camera, not all object categories are known to the ML model in advance since new objects of interest appear constantly in a user environment. As a result, it is important for object detection models to be continually learning -- they need to learn how to recognize new objects without suffering from the phenomenon of catastrophic forgetting, where the ML model forgets about old objects while learning about new ones. In this work, we discuss a new technology we have developed that can effectively do incremental learning for object detection in near-real time. We discuss the underlying mathematical framework of a novel loss function that enabled us to achieve state-of-the-art performance on benchmark datasets. We will also outline our efficient training and inference framework, which enabled our prototype system to successfully recognize objects in a real-world live demo scenario. We also discuss extensions of our incremental object detection work, where we can use auxiliary unlabeled data to get better models or use AutoML methods to automatically learn the best neural network architecture in the continuous learning mode. We next give a brief overview of a novel recurrent neural network model with attention that we have developed for the task of Visual Dialogue, where the user initiates a dialogue with the system regarding a picture. We conclude by discussing how incremental object detection, improved visual dialogue, and other novel research contributions form the cornerstones of a new framework of Conversational Vision, which is an active computer vision technology at the intersection of Natural Language Processing, Dialogue Understanding and Computer Vision.

avatar for Shalini Ghosh, Samsung Research America

Shalini Ghosh, Samsung Research America

Director of AI Research, Samsung Research America
Dr. Shalini Ghosh is the Director of AI Research at the Artificial Intelligence Center of Samsung Research America, where she leads a group working on Situated AI and Multi-modal Learning (i.e., learning from computer vision, language, and speech). She has extensive experience and... Read More →

Thursday January 24, 2019 2:40pm - 3:00pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Advancing State-of-the-art Image Recognition with Deep Learning on Hashtags
At Facebook everyday hundreds of millions of users interact with billions of visual contents. By understanding what's in an image, our systems can help connect users with the things that matter most to them. To improve our recognition system, I will talk about two main research challenges: how we train models at the scale of billions, and how we improve the reliability of the model prediction. Since current models are typically trained on data that are individually labeled by human annotators, scaling up to billions is non-trivial. We solve the challenge by training image recognition networks on large sets of public images with user-supplied hashtags as labels. By leveraging weakly supervised pretraining, our best model achieved a record-high 85.4% accuracy on ImageNet dataset.

avatar for Yixuan Li, Facebook AI (Computer Vision Group)

Yixuan Li, Facebook AI (Computer Vision Group)

Research Scientist, Facebook AI (Computer Vision Group)
Yixuan Li is a Research Scientist at Facebook AI, Computer Vision Group. She leads the research effort on large-scale visual learning with high dimensional label space. Before joining Facebook, she obtained her PhD from Cornell University in 2017. Yixuan's research interests are in... Read More →

Thursday January 24, 2019 3:00pm - 3:20pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Deeper Things: How Netflix Leverages Deep Learning in Recommendations and Search
In this talk, we will provide an overview of Deep Learning methods applied to personalization and search at Netflix. We will set the stage by describing the unique challenges faced at Netflix in the areas of recommendations and information retrieval. Then we will delve into how we leverage a blend of traditional algorithms and emergent deep learning methods to address these challenges. We will conclude with a note on future directions we plan on pursuing at Netflix.

avatar for Aish Fenton, Netflix

Aish Fenton, Netflix

Research Manager, Netflix
Aish is a Research Manager at Netflix. He leads the machine learning research team there for recommender systems and search algorithms. Aish has over 22 years of experience at the intersection of mathematics and software engineering. Prior to Netflix, Aish headed data science at Opentable... Read More →
avatar for Sudeep Das, Netflix

Sudeep Das, Netflix

Senior Researcher, Netflix
Sudeep Das is a Senior Researcher at Netflix, where his main focus is on developing the next generation of machine learning algorithms to drive the personalization, discovery and search experience in the product. Apart from algorithmic work, he also takes a keen interest in data visualizations... Read More →

Thursday January 24, 2019 4:00pm - 4:20pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


5 Lessons for Improving Training Performance
Learn the best practices for performance analytics and maintenance of a deep learning system. As GPU technology continues to advance, the demand for faster data continues to grow. In deep learning, input pipelines are responsible for a complex chain of actions that ultimately feed data into GPU memory, including reading from storage and pre-processing data. These pipelines bring together multiple hardware systems—networking, CPUs, and storage—along with sophisticated software systems to drive the data movement and transformation.

We'll use results of TensorFlow benchmarking on V100 DGX-1s to highlight ways that overall performance is impacted by various components of the pipeline, and we'll share key ways to keep an end-to-end system highly performant over time.

avatar for Emily Watkins, Pure Storage

Emily Watkins, Pure Storage

Solution Architect, Pure Storage
Emily Watkins is a Solution Architect at Pure Storage. She helps companies streamline their data pipeline to help scale as their AI Projects grow from infancy to delivering significant outcomes for the business. Emily's background is in research, real-time analytics tools, and artificial... Read More →

Thursday January 24, 2019 4:20pm - 4:40pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


The Myth of the interpretable, Robust, Compact and High Performance Deep Neural Network
Most progress in machine learning has been measured according to gains in test-set accuracy on tasks like image recognition. However, test-set accuracy appears to be poorly correlated with other design objectives such as interpretability, robustness to adversarial attacks or training compact networks that can be used in resource constrained environments. This talk will ask whether it is possible to have it all, and more importantly how do we measure progress when we want to train model functions that fulfill multiple criteria.

avatar for Sara Hooker, Google

Sara Hooker, Google

Artificial Intelligence Resident, Google
Sara Hooker is Artificial Intelligence Resident at Google Brain doing deep learning research on model compression and reliable explanations of model predictions for black-box models. Her main research interests gravitate towards interpretability, model compression and security. In... Read More →

Thursday January 24, 2019 4:40pm - 5:00pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA
Friday, January 25


Compère Welcome - Deep Learning Stage
Our compère for the Deep Learning Stage today will be Alicia Kavelaars, Co-Founder and CTO at OffWorld.

avatar for Alicia Kavelaars, OffWorld

Alicia Kavelaars, OffWorld

Co-Founder and CTO, OffWorld
Alicia is Co-Founder and Chief Technology Officer at OffWorld Inc. She brings over 15 years of experience in the aerospace industry developing and successfully launching systems for NASA, NOAA and the Telecommunications industry. In 2015, Alicia made the jump to New Space to work... Read More →

Friday January 25, 2019 9:00am - 9:10am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Learned Video Compression
We present an algorithm for video coding, learned end-to-end for the low-latency mode. In this setting, our approach outperforms all standard video codecs across nearly the entire bitrate range. To our knowledge, this is the first ML-based method to do so. We propose a novel architecture for video compression which generalizes motion estimation to perform any learned compensation beyond simpler translations. Our architecture allows for joint compression of motion and residual and can dynamically trade-off between them. It is also able to model multiple flow fields in the same frame. We propose an ML-based spatial rate control, which allows or model to adaptively change the bitrate across space for each frame. For the same quality traditional codecs achieve up to 60% larger code.

avatar for Lubomir Bourdev, WaveOne

Lubomir Bourdev, WaveOne

Co-Founder & CEO, WaveOne
Lubomir Bourdev is a co-founder and the CEO of WaveOne, Inc., a startup focusing on video compression with deep learning. He is also a founding member of Facebook AI Research and he founded and led the Facebook AML Computer Vision team responsible for the image and video content recognition... Read More →

Friday January 25, 2019 9:10am - 9:25am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Using AI to Transform Informational Videos and Our Watching Behavior
Videos account for about 75% of the internet traffic and enterprises are increasingly using videos for various informational purposes, including training of customers, partners and employees, marketing and internal communication. However, most viewers do not have the patience to watch these videos end-to-end and our video watching experience has not evolved much in over a decade. We present an AI-based approach to automatically index videos in the form of a table-of-contents, a phrase cloud and a searchable transcript, which helps summarize the key topics in a video and lets viewers navigate directly to the topics of interest. We use a combination of visual classification, object detection, automated speech recognition, text summarization, and domain classification, and show the results achieved on a range of informational videos. We conclude with some thoughts on the promise of transforming how informational videos are consumed as well as open problems and future directions.

avatar for Manish Gupta, VideoKen

Manish Gupta, VideoKen

CEO & Co-Founder, VideoKen
Dr. Manish Gupta is the co-founder and CEO of VideoKen Inc., a video technology startup. He has served as the Vice President and Director of Xerox Research Centre India and has held various leadership positions with IBM, including that of Director, IBM Research - India and Chief Technologist... Read More →

Friday January 25, 2019 9:25am - 9:40am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


DNA of an AI Powered Robotic Workforce for Extreme Environments
As practical applications of AI emerge in industrial robotics, we are starting to realize the potential of highly autonomous robotic systems not only capable of performing AI driven tasks in simulation or structured environments, but out in the field and even in extreme environments. However, there are optimum solutions that do not require the use of deep reinforcement learning or other Machine Learning methodologies. What is the right balance between powerful implementations of AI and traditional automation control to achieve a highly autonomous robotic system that can operate in remote unforgiving locations? Is there a right DNA for an AI powered robotic workforce for extreme environments?

avatar for Alicia Kavelaars, OffWorld

Alicia Kavelaars, OffWorld

Co-Founder and CTO, OffWorld
Alicia is Co-Founder and Chief Technology Officer at OffWorld Inc. She brings over 15 years of experience in the aerospace industry developing and successfully launching systems for NASA, NOAA and the Telecommunications industry. In 2015, Alicia made the jump to New Space to work... Read More →

Friday January 25, 2019 9:40am - 9:55am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Deep Robotic Learning
Deep learning has been demonstrated to achieve excellent results in a range of passive perception tasks, from recognizing objects in images to recognizing human speech. However, extending the success of deep learning into domains that involve active decision making has proven challenging, because the physical world presents an entire new dimension of complexity to the machine learning problem. Machines that act intelligently in open-world environments must reason about temporal relationships, cause and effect, and the consequences of their actions, and must adapt quickly, follow human instruction, and remain safe and robust. Although the basic mathematical building blocks for such systems -- reinforcement learning and optimal control -- have been studied for decades, such techniques have been difficult to extend to real-world control settings. For example, although reinforcement learning methods have been demonstrated extensively in settings such as games, their applicability to real-world environments requires new and fundamental innovations: not only does the sample complexity of such methods need to be reduced by orders of magnitude, but we must also study generalization, stability, and robustness. In this talk, I will discuss how deep learning and reinforcement learning methods can be extended to enable real-world robotic control, with an emphasis on techniques that generalize to situations, objects, and tasks. I will discuss how model-based reinforcement learning can enable sample-efficient control, how model-free reinforcement learning can be made efficient, robust, and reliable, and how meta-learning can enable robotic systems to adapt quickly to new tasks and new situations.

avatar for Sergey Levine, UC Berkeley

Sergey Levine, UC Berkeley

Assistant Professor, UC Berkeley
Sergey Levine received a BS and MS in Computer Science from Stanford University in 2009, and a Ph.D. in Computer Science from Stanford University in 2014. He joined the faculty of the Department of Electrical Engineering and Computer Sciences at UC Berkeley in fall 2016. His work... Read More →

Friday January 25, 2019 9:55am - 10:15am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Control Algorithms for Imitation Learning from Observation
Imitation learning is a paradigm that enables autonomous agents to capture behaviors that are demonstrated by people or other agents. Effective approaches, such as Behavioral Cloning and Inverse Reinforcement Learning, tend to rely on the learning agent being aware of the low-level actions being demonstrated.  However, in many cases, such as videos or demonstrations from people (or any agent with a different morphology), the learning agent only has access to observed state transitions.  This talk introduces two novel control algorithms for imitation learning from observation: Behavioral Cloning from Observation (BCO) and Generative Adversarial Imitation from Observation (GAIfO).

avatar for Peter Stone, University of Texas

Peter Stone, University of Texas

Founder & Director of the Learning Agents Research Group, University of Texas
I am the founder and director of the Learning Agents Research Group (LARG) within the Artificial Intelligence Laboratory in the Department of Computer Science at The University of Texas at Austin, as well as associate department chair and chair of the University's Robotics Portfolio... Read More →

Friday January 25, 2019 10:45am - 11:05am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Meta-Learning Deep Networks
Deep learning has enabled significant advances in a variety of domains; however, it relies heavily on large labeled datasets. I will discuss how we can use meta-learning, or learning to learn, to enable us to adapt deep models to new tasks with tiny amounts of data, by leveraging data from other tasks. By using these meta-learning techniques, I will show how we can build better unsupervised learning algorithms, build agents that can adapt online to changing environments, and build robots that can interact with a new object by watching a single demonstration.

avatar for Chelsea Finn, Google Brain & Berkeley AI Research

Chelsea Finn, Google Brain & Berkeley AI Research

Research Scientist & Post-doctoral Scholar, Google Brain & Berkeley AI Research
Chelsea Finn is a research scientist at Google Brain and post-doctoral scholar at Berkeley AI Research. Starting in 2019, she will join the faculty in CS at Stanford University. She is interested in how learning algorithms can enable machines to acquire general notions of intelligence... Read More →

Friday January 25, 2019 11:05am - 11:30am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Latent Structure in Deep Robotic Learning
Traditionally, deep reinforcement learning has focused on learning one particular skill in isolation and from scratch. This often leads to repeated efforts of learning the right representation for each skill individually, while it is likely that such representation could be shared between different skills. In contrast, there is some evidence that humans reuse previously learned skills efficiently to learn new ones, e.g. by sequencing or interpolating between them.
In this talk, I will demonstrate how one could discover latent structure when learning multiple skills concurrently. In particular, I will present a first step towards learning robot skill embeddings that enable reusing previously acquired skills. I will show how one can use these ideas for multi-task reinforcement learning, sim-to-real transfer and imitation learning.

avatar for Karol Hausman, Google Brain

Karol Hausman, Google Brain

Research Scientist, Google Brain
Karol Hausman is a Research Scientist at Google Brain in Mountain View, California working on robotics and machine learning. He is interested in enabling robots to autonomously acquire general-purpose skills with minimal supervision in real-world environments. His current research... Read More →

Friday January 25, 2019 11:30am - 11:50am
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Applying Multimodal Integrated Learning Behavioral Analysis and Virtual Personalized Assistant to Adaptive Education
Summary, in this talk, Dr. Edgar Kalns who is the lead researcher on the SRI - Squirrel AI Joint Lab will discuss two exciting AI and Machine Learning applications for adaptive education. He will illustrate the theoretical foundation, technical approach, and research findings from the MILBA and VPA projects.

avatar for Edgar Kalns, Squirrel AI

Edgar Kalns, Squirrel AI

Director of SRI-Squirrel AI Joint Collaboration, Squirrel AI
Edgar Kalns, Ph.D., is leading SRI’s Studio on Aging, an institute-wide initiative to help seniors live independent and fulfilled lives, simultaneously addressing the widening caregiver gap. The Studio's goal is to create world-changing interdisciplinary solutions by fusing technologies... Read More →

Friday January 25, 2019 11:50am - 12:10pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Brand is Beyond Logos – Understanding Visual Brand
Logos come to mind when we think about iconic brands. However, a spectrum of visual cues is used to establish the signature of a brand. This includes color, pattern, shape. We train deep neural network to predict a variety of fashion brand and analyze visual representations using strength and extent of neuron activations. Logo is demonstrated to be at one end of the spectrum. Study of versatility of neurons shows that they are diverse in nature and contain specialists and generalists. Potential applications of making neural network explainable include personalization, elimination of bias in prediction, model improvement.

avatar for Robinson Piramuthu, eBay

Robinson Piramuthu, eBay

Chief Scientist for Computer Vision, eBay
Robinson Piramuthu joined eBay in 2011 and is currently the Chief Scientist for Computer Vision. He has over 20 years of experience in computer vision which includes large scale visual search, coarse and fine-grained visual recognition, object detection, computer vision for fashion... Read More →

Friday January 25, 2019 12:10pm - 12:30pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Applying Deep Learning to Article Embedding for Fake News Evaluation
In this talk we explore real world use case applications for automated “Fake News” evaluation using contemporary deep learning article vectorization and tagging. We begin with the use case and an evaluation of the appropriate context applications for various deep learning applications in fake news evaluation. Technical material will review several methodologies for article vectorization with classification pipelines, ranging from traditional to advanced deep architecture techniques. We close with a discussion on troubleshooting and performance optimization when consolidating and evaluating these various techniques on active data sets.

avatar for Mike Tamir, Uber ATG

Mike Tamir, Uber ATG

Head of Data Science, Uber ATG
Mike serves as Head of Data Science at Uber ATG, UC Berkeley Data Science faculty, and Director of Phronesis ML Labs. He has led teams of Data Scientists in the bay area as Chief Data Scientist for InterTrust and Takt, Director of Data Sciences for MetaScale/Sears, and CSO for Galvanize... Read More →

Friday January 25, 2019 1:15pm - 1:40pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Applying Deep Learning To Airbnb Search
Searching for homes is the primary mechanism guests use to find the place they want to book at Airbnb. The goal of search ranking is to find guests the best possible options while rewarding the most deserving hosts. Ranking at Airbnb is a quest to understand the needs of the guests and the quality of the hosts to strike the best match possible. Applying machine learning to this challenge is one of the biggest success stories at Airbnb. Much of the initial gains were driven by a gradient boosted decision tree model. The gains, however, plateaued over time. This talk discusses the work done in applying neural networks in an attempt to break out of that plateau. The talk focuses on the elements we found useful in applying neural networks to a real life product. To other teams embarking on similar journeys, we hope this account of our struggles and triumphs will provide some useful pointers. Bon voyage!

avatar for Malay Haldar, Airbnb

Malay Haldar, Airbnb

Machine Learning Engineer, Airbnb
Malay is a machine learning engineer working on search ranking at Airbnb. Prior to Airbnb, Malay worked on applying machine learning to Google Play search with the goal of understanding the functionality of each app. Before machine learning, Malay worked on web-scale infrastructure... Read More →

Friday January 25, 2019 1:40pm - 2:00pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


End-to-End Conversational System for Customer Service Application
Goal-oriented conversational systems are typically multi-turn, relying on the entire conversation thus far to generate a response to user input. Many of these systems use dialog state tracking or belief tracking, to either rank candidate responses from a pool of templates or generate responses directly while others are end-to-end. End-to-end models for goal-oriented conversational systems have become an increasingly active area of research.
In this talk, I will present our recent efforts to build end-to-end conversational models for customer service application. We use historical chat transcripts and customer profile data to build models, and test with live customers using a human-in-the-loop research platform. We experiment with sequence-to-sequence model that generates responses word by word, and multi-encoder based ranking model to score template responses. I will compare these approaches as they apply to customer service domain.

avatar for Manisha Srivastava, Amazon Customer Service

Manisha Srivastava, Amazon Customer Service

Machine Learning Scientist, Amazon Customer Service
Manisha Srivastava is a machine learning scientist at Amazon Customer Service, working to improve the customer experience using NLP techniques. Prior to Amazon she worked at TripAdvisor, focusing on using machine learning to improve the quality of business listings. She received her... Read More →

Friday January 25, 2019 2:00pm - 2:20pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


Human and Artificial Intelligence in Healthcare
The combination of breakthroughs in AI and Machine Learning and increasing amounts of digitized medical data have generated significant excitement about the potential to automate medical decision making processes. Among these, there are significant opportunities in designing solutions for settings where AI and ML systems can work seamlessly with human experts to provide more efficient and accurate patient care. In this talk, I outline one such problem, that of medical expert disagreement. We study the application of machine learning to predict patient cases which are likely to give rise to maximal expert disagreement. We show one can develop and train AI models to predict an uncertainty score for a patient, identifying cases where large disagreements ensue, and flagging that patient for a medical second opinion. Methodologically, we formalize the importance of doing direct prediction of these uncertainty scores, instead of a two step process of diagnosis and postprocessing, evaluating on a  gold-standard adjudicated dataset.

avatar for Maithra Raghu, Google Brain/Cornell University

Maithra Raghu, Google Brain/Cornell University

Research Scientist, Google Brain/Cornell University
Maithra Raghu is a PhD Candidate in Computer Science at Cornell University, and a Research Scientist at Google Brain. Her research interests are in developing principled tools to empirically study the representational properties of deep neural networks, and apply these insights to... Read More →

Friday January 25, 2019 2:20pm - 2:40pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA


A Deep Learning Model for Early Prediction of the Diagnosis of Alzheimer Disease from 18F-FDG PET scan of the Brain
Over 5 million Americans are affected by Alzheimer's disease and have cost over 200 billion dollars in direct and indirect costs . Early and accurate diagnosis of Alzheimer's disease is important because it opens the possibility of therapeutic intervention to slow or halt the disease progression. Unfortunately, Alzheimer disease (AD) remains a diagnosis based on clinical grounds and most diagnosis gets established at a late stage when too many neurons have been lost. Much research effort has been made on biochemical and imaging tests
to improve our early diagnostic capability but most have met with mild to moderate accuracy. 18F-FDG PET scans of the brain utilize radioactive glucose to image the energy uptake pattern in various parts of the brain, which has recently been implicated to change in subtle ways in Alzheimer's disease. In this study, we develop and validate a deep learning algorithm that predicts the final diagnosis of Alzheimer disease (AD), mild cognitive impairment, or neither, approximately 6 years before the final diagnosis from these PET scans.

Prospective 18F-FDG PET brain images from the Alzheimer's Disease Neuroimaging Initiative (ADNI) (2109 imaging studies from 2005 to 2017, 1002 patients) and retrospective independent test set (40 imaging studies from 2006 to 2016, 40 patients) were collected. Final clinical diagnosis at follow-up was recorded. Convolutional neural network of Inception architecture was trained on 90% of ADNI data set and tested on the remaining 10%, as well as the independent test set, with performance compared to radiologic readers. Model was analyzed with sensitivity, specificity, receiver operating characteristic (ROC), saliency map, and t-distributed stochastic neighbor embedding. The algorithm achieved area under the ROC curve of 0.98 (95% confidence interval: 0.94, 1.00) when evaluated on predicting the final clinical diagnosis of AD in the independent test set (82% specificity at 100% sensitivity), an average of 75.8 months prior to the final diagnosis, which in ROC space outperformed reader performance (57% [four of seven] sensitivity, 91% [30 of 33] specificity; P < .05). Saliency map demonstrated attention to known areas of interest but with focus on the entire brain.

By using fluorine 18 fluorodeoxyglucose PET of the brain, a deep learning algorithm developed for early prediction of Alzheimer disease achieved 82% specificity at 100% sensitivity, an average of 75.8 months prior to the final diagnosis.

avatar for Jae Ho Sohn, UCSF Medical Center

Jae Ho Sohn, UCSF Medical Center

Radiology Resident, UCSF Medical Center
Jae Ho Sohn, MD, MS is a radiology resident at UCSF Medical Center. As a physician with engineering background, his research focuses on the intersection of big data and radiology. He and his team has been working on a number of computer vision and natural language processing algorithms... Read More →

Friday January 25, 2019 2:40pm - 3:00pm
Grand Ballroom Hyatt Regency San Francisco, 5 Embarcadero Center, San Francisco, CA 94111, USA