20180131-DQN-keras-refine.ipynb
-
Objective
- Find the minimum of an unknown function
-
Toolkit
- keras >= 2.0
- python 2.7, but I guess 3.5 is workable
- matplotlib, numpy
-
Environment
-
Agent
- An agent with Deep Q-learning is applied
-
Result
- Loss is not stable
-
Future Work
- Use policy gradient
- Modify reward function