Details

Agenda:

17:30-18:30: Value-based Deep Reinforcement Learning by Tom Zahavy

18:30-18:40: Networking break

18:40-19:30: Deep Learning & Genetic Algorithms by Yam Peleg

Value-based Deep Reinforcement Learning by TOM Zahavy

In this talk, Tom will cover recent advances in value-based Deep Reinforcement Learning, from foundations to state-of-the-art, including three of his papers on that topic. More details on Tom’s homepage:

http://tomzahavy.wixsite.com/zahavy

Tom Zahavy is a PhD student at the Technion, working with Prof. Shie Mannor on Deep Reinforcement Learning. His research focuses on learning value functions with deep neural networks and is closely related to the Deep Q Network (DQN). In particular, Tom worked on visualizing and interpreting DQN's representation (ICML16), improving DQN by combining it with linear function approximation (NIPS17), and extending the DQN architecture to support temporally extended actions and hierarchy (AAAI17). Other than that, Tom also worked on machine and deep learning applications for real-world problems including e-commerce, signal processing, and physics.
Over the course of his PhD Tom interned at Walmart.com where he worked on multi-modal classification for e-commerce, Facebook AI Research where he worked on transfer learning for deep reinforcement learning with GANs and Google research where he is currently working on theory for value functions decomposition.

Deep Learning & Genetic Algorithms by Yam Peleg

Deep Learning networks are a type of models that discoversimportant features. These networks determine featureswithout supervision, and are adept at learning high level abstractionsabout their data sets. This talk examines the use of Genetic Algorithms as a trainingmechanism for Deep Learning networks, with emphasis on trainingnetworks with a large number of layers, each of which is trained independentlyto reduce the computational burden and increase the overallflexibility of the algorithm.

Yam Peleg is the founder of Deep Trading ltd. He was a quantitative high frequency trader for more than four years and deep trading is his second startup. He is also a major contributor to the python community who spoke at python conferences around the world, including PyCon, PyData ,SciPy and more.