Decoupling feature extraction from policy learning: assessing benefits of state representation learning in goal based robotics

January 24, 2019 · Entered Twilight · 🏛 arXiv.org

"Last commit was 7.0 years ago (≥5 year threshold)"

Evidence collected by the PWNC Scanner

Repo contents: .coveragerc, .gitignore, LICENSE, README.md, configs, data, docs, environment.yml, evaluation, imgs, logs, losses, models, pipeline.py, plotting, preprocessing, run_tests.sh, server.py, srl_baselines, tests, train.py, utils.py

Authors Antonin Raffin, Ashley Hill, René Traoré, Timothée Lesort, Natalia Díaz-Rodríguez, David Filliat arXiv ID 1901.08651 Category cs.LG: Machine Learning Cross-listed cs.RO, stat.ML Citations 62 Venue arXiv.org Repository https://github.com/araffin/srl-zoo ⭐ 163 Last Checked 3 months ago

Abstract

Scaling end-to-end reinforcement learning to control real robots from vision presents a series of challenges, in particular in terms of sample efficiency. Against end-to-end learning, state representation learning can help learn a compact, efficient and relevant representation of states that speeds up policy learning, reducing the number of samples needed, and that is easier to interpret. We evaluate several state representation learning methods on goal based robotics tasks and propose a new unsupervised model that stacks representations and combines strengths of several of these approaches. This method encodes all the relevant features, performs on par or better than end-to-end learning with better sample efficiency, and is robust to hyper-parameters change.