Summary of Investigating Pre-training Objectives For Generalization in Vision-based Reinforcement Learning, by Donghu Kim et al.
Investigating Pre-Training Objectives for Generalization in Vision-Based Reinforcement Learning
by Donghu Kim, Hojoon Lee, Kyungmin Lee, Dongyoon Hwang, Jaegul Choo
First submitted to arxiv on: 10 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper introduces a new benchmark for evaluating pre-training methods in vision-based Reinforcement Learning (RL). The Atari Pre-training Benchmark (Atari-PB) consists of 10 million transitions from 50 Atari games, which are used to train a ResNet-50 model. The authors evaluate the generalization ability of different pre-training objectives across diverse environment distributions. They find that task-agnostic features improve generalization, while task-specific knowledge enhances performance in similar environments but not varied ones. The study contributes to understanding the effectiveness of various pre-training methods and their limitations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about a new way to test how well computers can learn to play games using pictures. It creates a big dataset with 10 million examples from 50 different Atari games, then trains a computer model on this data. The goal is to see if the model can work well in different situations, like playing a game that’s very similar or very different from one it was trained on. The results show that some ways of training the model are better than others at doing this. This helps us understand how to make computers better at learning new tasks. |
Keywords
» Artificial intelligence » Generalization » Reinforcement learning » Resnet
