A Probabilistic Model for Skill Acquisition with Switching Latent Feedback Controllers

This repo tries to learn to segment skills from image demonstrations and learn a controller for each skill. Each controller has a goal and gain as its own private property.

Model Architecture

Project Website

https://sites.google.com/view/skillacquisition/home

Job

The following code starts a job using a toy dataset.

python -u train.py -m "test"