authors: Aaron Dupont, Blake Allen, Harry Ly, Christopher Schayer, Matthew Guidry, Wilson Zhu
purpose: the following project is to teach our AI how to operate a bipedal-walker
- setup a conda env with python3.4 and activate
- create conda environment -> conda create -n py34 python=3.4
- install pybox2d thru conda -> conda install -c https://conda.anaconda.org/kne pybox2d
- install openai's gym thru pip -> pip install --user gym
- install neat thru pip -> pip install --user neat-python
- install graphviz thru pip -> pip install --user graphviz
- also install the graphviz application using the standard application manager
- install matplotlib thru pip -> pip install --user matplotlib
- how to run -> python main.py
- optional arguments:
- run in parallel: -t [# of threads] note: this largely depends on how many cores your cpu has
- run from checkpoint: -f [path of the checkpoint file] note: you need this if you want to use the replay functionality
- render: -m replay
abstract: due Oct. 9thsetup neural networkrename this repo to something elseformat Chris's ugly codetweak neural network to pass conditionwrite a logger for the programadd a visual representation of the neural network- ~~start experimental phase (compare to other neural networks) ~~
write up reportmake powerpoint slides for presentation (reserved for Aaron, Blake, and Harry)