Introduction to Game Training using Deep RL
From AlgphaGo to VizDoom, Deep Reinforcement Learning has revolutionarized the way in which we learn game environments. Especially for people playing CS and Dota in Colleges, having a smarter bot by thier side can help them sleep well, while the bots are fighting against each other.
In the talk, we will discussing intuition behind design of Alphago. We will also discuss, what makes Reinforcement Learning Unique as compared to supervised Learning counterparts. The presentation will end with a small 10 second video of trained game.
Now the topics I am planning to cover are :
1. Why Deep Reinforcement Learning and not Deep Learning for games?
2. How did AlphaGo beat humans in game of Go.
3. Superbasic intuition of RL
4. What are the state of the art approaches for training
- Deep RL Approaches : Policy Gradient Approaches, Actor Critic Methods
- Implementation Frameworks : Unity-ML, OpenAI Baselines
5. Difficulties with Deep RL approaches
6. How can you start training your favourite game?
Basics of Neural Networks
Jaley is an alumni of Video Analytics Lab,CDS at IISc (Indian Institute of Science) and has extensively worked on Deep Learning applications. He had developed Expresso, a tool for Deep Learning, keeping in eye the problem faced by Deep Learning beginners in 2013-14, when there was no Tensorflow, DIGITS or Keras.
He has worked on Spark and Hadoop for predictive modelling and temporal aspects of recommender systems during his stay at Vizury, which is an ad re-targeting company. One of his significant contribution at Vizury was sustained improvement in the recommender systems on both web as well as app platform. He had also worked in Samsung-Harman as Senior Data-scientist in music domain and autonomous parking problem.
Currently, he is working on youtube channel Crazymuse, where intends to simplify the abstract mathematical intutions in Research papers.
Above all, he is an amazing gujju cook and loves South-Indian music especially when rendered by Jayshree Ramanath.