Is Augmentation Actually Effective In Reinforcement Learning?
Three main points
✔️ State-of-the-Art results with DeepMind Control and OpenAI Gym
✔️ Augmentation makes it more versatile in testing
✔️ New Augmentation Methodology
Reinforcement Learning with Augmented Data
written by Michael Laskin, Kimin Lee, Adam Stooke, Lerrel Pinto, Pieter Abbeel, Aravind Srinivas
(Submitted on 30 April 2020)
Comments: First two authors contributed equally, website: this https URL code: this https URL and this https URL
Subjects: Machine Learning (cs.LG), Machine Learning (stat.ML)
Introduction
Image as input Reinforcement learning has various problems, such as poor data efficiency and lack of versatility. In this article, we introduce a paper called Reinforcement Learning with Augmented Data (RAD), which improves learning efficiency by performing augmentation on input images. Although it is often used and the method is very simple, it has been found to be effective in augmenting previously unused images and has successfully improved the efficiency of training many RL algorithms.
Here's how we've increased learning efficiency, and versatility when testing.
To read more,
Please register with AI-SCHOLAR.
ORCategories related to this article