Hacker News Comments on
Reinforcement Learning 10: Classic Games Case Study
DeepMind
·
Youtube
·
41
HN points
·
1
HN comments
- This course is unranked · view top recommended courses
Hacker News Stories and Comments
All the comments and stories posted to Hacker News that reference this video.⬐ MasterScratThis video is part of the "Advanced Deep Learning & Reinforcement Learning" series: https://www.youtube.com/playlist?list=PLqYmG7hTraZDNJre23vqC...> This course, taught originally at UCL and recorded for online access, has two interleaved parts that converge towards the end of the course. One part is on machine learning with deep neural networks, the other part is about prediction and control using reinforcement learning. The two strands come together when we discuss deep reinforcement learning, where deep neural networks are trained as function approximators in a reinforcement learning setting.
> The deep learning stream of the course will cover a short introduction to neural networks and supervised learning with TensorFlow, followed by lectures on convolutional neural networks, recurrent neural networks, end-to-end and energy-based learning, optimization methods, unsupervised learning as well as attention and memory. Possible applications areas to be discussed include object recognition and natural language processing.
There is a fascinating (and charming) paper by Alan Turing that describes his "Turochamp" chess 'engine'. Apparently, it was the first program capable of playing a complete game of chess, and the first program that could be described as a computer game (although it sadly only ever existed on paper). The general pattern he outlines (a heuristic evaluation function with hand-tuned weights, along with minimax game tree search--i.e. backwards induction) has formed the basis of most chess engines, both ancient and modern. Here's the original copy: https://docs.google.com/file/d/0B0xb4crOvCgTNmEtRXFBQUIxQWs/...Intriguingly, Turing posed the question, "Could one make a machine to play chess, and to improve its play, game by game, profiting from its experience?" This reinforcement learning approach to chess did not enjoy much success--until AlphaZero. That story that has been well-told in many places, but perhaps best so by David Silver in this recently released lecture by DeepMind: https://www.youtube.com/watch?v=ld28AU7DDB4. The first ~40 mins are a lucid explanation of the classical methods, and the rest covers RL/MCTS/AlphaZero.
⬐ opo>..."Could one make a machine to play chess, and to improve its play, game by game, profiting from its experience?" This reinforcement learning approach to chess did not enjoy much success--until AlphaZero.Don't forget Samuel's computer checkers program from 1959. It was among the world's first successful self-learning programs.