Getting to Know The Elephant: Real-Time Debugging and Visualization for Deep Learning
Who is this presentation for?Software Engineer, Machine Learning Engineer, Deep Learning Researcher
While deep learning has been wildly successful at many tasks, our understanding of why these models work or don’t have been severely limited. Simultaneously, the area of model understanding and explaining its success of failures has been fairly under-represented area. In this talk we present a new tool we call TensorWatch built at Microsoft Research to help and accelerate debugging and visualization for deep learning. This tool has been designed from the ground up rethinking fundamental primitives needed for such tasks and putting composability front and center. Our framework is real-time, interactive and provides accelerated insights on how deep learning model works. We will walk through different aspects of workflows that machine learning practitioners frequently encounter. For example, quiqly answering questions such as what is happening in my training, is my model architecture good enough, why these my model is failing for this given data? This presentation will include live interactive code running for demos.
Prerequisite knowledgeDeep learning basics, Python programming, Jupyter Notebook, Statistics and Deep Learning metrics 101
What you'll learn
Shital Shah is Principal Research Engineer at Microsoft Research AI. His interests include deep learning debugging, visualization, simulation, autonomous vehicles and robotics. He has been working at Microsoft for past 12 years architecting, designing and developing large scale machine learning systems that heavily leverages distributed computing. He has contributed in research and engineering in various roles at Microsoft including technical lead, architect, engineering manager and research engineer. More recently at Bing, he founded and lead the team to design and develop distributed machine learned clustering platform for web-scale data. At Microsoft Research, he conceived and lead the development of AirSim, a physically and visually realistic cross platform simulator for AI research as test bed for designing, applying and evaluating reinforcement learning algorithms for the smart agents in open dynamic world.
Leave a Comment or Question
Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?
Join the conversation here (requires login)
For conference registration information and customer service
For more information on community discounts and trade opportunities with O’Reilly conferences
For information on exhibiting or sponsoring a conference
View a complete list of Strata Data Conference contacts