This is a simple example of reinforcement learning. Hope this encourages you to build something interesting yourself.
Change from iter2. Change from iter1.
The environment is a 1x7 grid. The agent starts at S, with each
episode ending at G and the
available actions are left or right. There are three different
rewards systems used. Each .py file implements different
reward.
All the available code is in SliderReward[1-3].py files. Each file contains complete and independant code.
Note: numpy should be installed.
To run any of the files:
python3 SliderReward[1-3].py <no-of-episodes>Here <no-of-episodes> is the amount of iterations for training
the agent. If no argument is provided, then the default is 50.
Example
$ python3 SliderReward1.py
$ python3 SliderReward1.py 100