Abstract

This talk provides an introduction to Deep Reinforcement Learning. First we present Reinforcement Learning. Both Q-Learning and Temporal Difference Learning are discussed. We  provide an accelerated introduction to Artificial Neural Nets and specifically the Backpropagation algorithm. We then cover  Deep Learning topics and describe Convolutional Neural Nets and specifically the concepts of local receptive fields, feature maps and max-pooling.  Next we explain how Reinforcement Learning has  been successfully combined with Deep Learning. Finally, time permitting we will describe how Google DeepMind used Deep Reinforcement Learning to play and master Atari games and how the AlphaGo program beat world champion Go player Lee Se-dol earlier this year. Code examples will be provided in Clojure. 

References

• Reinforcement Learning  

–http://personal.disco.unimib.it/Vanneschi/McGrawHill_-_Machine_Learning_-Tom_Mitchell.pdf
–https://webdocs.cs.ualberta.ca/~sutton/book/ebook/the-book.html

• Neural Nets & Deep Learning

–http://personal.disco.unimib.it/Vanneschi/McGrawHill_-_Machine_Learning_-Tom_Mitchell.pdf ;
 –http://neuralnetworksanddeeplearning.com/chap2.html
- http://deeplearning.net/tutorial/deeplearning.pdf

 • Convolutional Neural Networks
 –http://cs231n.github.io/convolutional-networks/
 - http://neuralnetworksanddeeplearning.com/chap6.html

• Deep Reinforcement Learning –http://www0.cs.ucl.ac.uk/staff/d.silver/web/Resources_files/deep_rl.pdf ;

 •AlphaGo
– https://storage.googleapis.com/deepmind-media/alphago/AlphaGoNaturePaper.pdf
– http://deeplearningskysthelimit.blogspot.com/

Bio

Pierre de Lacaze has over 20 years experience with Lisp and AI based technologies. He holds a Bachelor of Science in Applied Mathematics and Computer Science and a Master’s Degree in Computer Science. He is the President of  LispNYC and Director of Machine Intelligence at Shareablee.