Demystifying deep reinforcement learning
Details (date, time, venue) and tickets for this workshop are available here: https://hasgeek.com/anthillinside/deep-reinforcement-learning-workshop/
Deep reinforcement learning combines reinforcement learning algorithms with deep learning paradigm. This hands-on workshop aims to familiarize the attendees with deep RL by working through notebook examples. Apart from the core concepts, we will also discuss recent advances, and practical implementations of deep RL.
Introduction to reinforcement learning
- how it is different than supervised and unsupervised learning
- basics of reinforcement learning
- controller/agent vs input vs environment/experiment vs reward
- MDP problem, MAB problem and other such variations
Deep reinforcement learning along with notebook examples: learning to play mario/alphago/a-game via Deep Q learning
- what is Q learning
- how to train a deep-learning-based-controller as part of Q learning
- advantages and limitations of Q learning
deep reinforcement learning along with notebook examples: learning to play the same game via proximal policy gradient
- what is proximal policy gradient
- how to train a deep-learning-based-controller as part of proximal policy optimization
- advantages and limitations of PPO
recent advances in deep reinforcement learning
- neural architecture search
- data distribution search
- limitations of deep reinforcement learning - cost, time, utility
When and how to apply deep reinforcement learning in your work
- stock trading
- recommendation systems
- dialog systems
- autonomous driving
Participants should bring their own laptops with pytorch + python 3.5.
- Uma Sawant is a Sr. machine learning engineer in the Artificial Intelligence group at Linkedin, Bangalore. She holds a PhD in computer science from IIT Bombay. Prior to joining for PhD, she worked as a research engineer in Yahoo research labs, Bangalore. She is a recipient of Google India Women in Engineering award, 2008. She has authored multiple papers in top tier conferences such as KDD, WWW, EMNLP.
- Vijay Gabale is cofounder and CTO of Infilect. Prior to cofounding Infilect, Vijay was a research scientist with IBM research. Vijay has published research papers in top tier conferences such as SIGCOMM, KDD and has several patents to his name. He holds a PhD in Computer Science from IIT Bombay.