An implementation of Locally Weighted Projection Regression

Readme for submission of Imitation Learning Seminar:

Included files:

• HS-IL-Jupyter-Notebook-Peter_Muschick (Folder containing jupyter notebook)
• _README.txt (current file)
• full_dataset_120k.csv (Dataset containing 120k lines of training data, created by harsha_evolution.py)
• harsha_evolution.py (Evolutionary algorithm used to create training data)
• harsha_evolution_cropped.csv (Dataset containing 4,8k lines of training data, created by harsha_evolution.py, only first 100 timestop used)
• HS-IL-Presentation-Peter_Muschick.pdf (file containing the presentation)
• linear.csv (test file containing points in linear order)
• linear_plateau.csv (test file containing points in linear order with a short pleateau)
• lwpr_algorithm.py (lwpr algorithm file)
• main.py (The main python script containing the cartpole open gym environment)
• networking.py (contains the UDP networking part for main.py)
• sharvar_keras.py (generates training data with an Proximal Policy Optimization algorithm)
• sharvar_keras_data.csv (created data from sharvar_keras.py)
• sinus_noise.csv (creates scattered sinus points)

Remarks/explanation to specific files:

main.py. - This file needs to be executed with python 3.x (open gym is only running with python 3.x)

lwpr_algorithm.py - https://github.com/jdlangs/lwpr - Check their README.txt - Basically a python 2.7 interpreter (32 bit) with a few packages (described in their README.txt) and GCC installed on top of it needs to run it