HemuManju / ahead

AtariHead exploration for Inverse Reinforcement Learning

ahead

A short description of the project.

• Exploration and modelling of AtariHead dataset for Inverse Reinforcement Learning

Data Preparation

• Run python src/data/data_setup.py
  • Downloads the data into src/data/raw
  • Creates interim data formats src/data/interim
  • Creates Processed data src/data/processed - TODO ( The models below can still be run without this.)
  • Does other structural setups

Gaze prediction

• CNN Model at src/models/cnn_gaze.py
  • Run with src/features/gaze_pred.py
• MDN model at src/models/mdn.py - In process

Supervised action learner

• Model at src/models/action_sl.py
  • Run with src/features/act_pred.py

Reinforcement Learning

• TODO

Inverse Reinforcement Learning

• TODO

Project based on the cookiecutter data science project template. #cookiecutterdatascience