Advanced Machine Learning Project
Topic
- Reinforcement learning
Member
- Ethan Li
- Tianchen Wang
- Yuki Kitayama
- Liwei Zhang
General idea about reinforcement learning
-
Make environment (state space) with reward and actions that an agent take (action space), set a goal that the agent wants to achieve, and estimate the optimal path by model or algorithm.
-
Brainstorming
- Campus walk
- Make a miniature Columbia campus. Agent starts from a building, buy and take out a lunch from a food truck or deli, and find an optimal location to eat it.
-
Use open environment from OpenAI
- The Open ai Gym contains multiple environments that we can work on.Such as invader and Black Jack. So maybe we can take a look at the website and find something interesting to work on.https://gym.openai.com/envs/#atari
(Write your idea) (Explain your idea) Challenging part
- Mapping the environment and specify all the possible moves (This equals to making Q table if you apply Q-learning). Theoretical things
- Q-learning
- Markov decision process
- Model for decision making, which you can use when some part is random, but you know some part.
- A Markov reward process with a decision factor in it.
- SARSA (State Action Reward next State and next Action) Very similar to Q-learning, but the difference is that Q-learning takes argmax, but SARSA doesn’t.
Resource (Reference)
-
Reinforcement Learning - With Open AI, TensorFlow and Keras Using Python; Abhishek Nandy, Manisha Biswas; apress
-
Reinforcement Q-learning from scratch in python with OpenAI gym https://www.learndatasci.com/tutorials/reinforcement-q-learning-scratch-python-openai-gym/
-
Reinforcement learning methods and tutorials (Github repo) https://www.kaggle.com/charel/learn-by-example-reinforcement-learning-with-gym (A kaggle example) https://github.com/MorvanZhou/Reinforcement-learning-with-tensorflow
-
莫煩 python 教學 (Youtube about RL) https://www.youtube.com/channel/UCdyjiB5H8Pu7aDTNVXTTpcg
https://github.com/openai/gym/blob/master/docs/creating-environments.md
https://github.com/openai/gym-soccer
How to modify Taxi environment
- Open taxi.py. It contains map, so we modify this map, and adjust number of states, columns, rows, and so on.