Summary of Adversarial Imitation Learning Via Boosting, by Jonathan D. Chang et al.
Adversarial Imitation Learning via Boosting
by Jonathan D. Chang, Dhruv Sreenivas, Yingbing Huang, Kianté Brantley, Wen Sun
First submitted to arxiv on: 12 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 Medium Difficulty summary: Adversarial Imitation Learning (AIL) has been a dominant framework in various imitation learning applications. The Discriminator Actor Critic (DAC) algorithm has demonstrated effectiveness in improving sample efficiency and scalability to higher-dimensional observations. However, the original AIL objective is on-policy, and DAC’s off-policy training does not guarantee successful imitation. This paper presents a novel and principled AIL algorithm via boosting, called AILBoost, which maintains an ensemble of properly weighted weak learners (policies) and trains a discriminator that witnesses the maximum discrepancy between the distributions of the ensemble and the expert policy. The algorithm uses a weighted replay buffer to represent the state-action distribution induced by the ensemble, allowing discriminators to be trained using the entire data collected so far. Empirically, AILBoost outperforms DAC on both controller state-based and pixel-based environments from the DeepMind Control Suite. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: This paper is about improving a type of artificial intelligence called imitation learning. Imitation learning helps machines learn new tasks by copying what they see others do. The current best way to do this, called Discriminator Actor Critic (DAC), has some limitations. In this research, scientists created a new algorithm that can learn even better and more efficiently. This new algorithm is called AILBoost. It’s like a team of smaller learners working together to figure out what the expert does best. The researchers tested their new algorithm on different types of tasks and found it was better than DAC in most cases. |
Keywords
* Artificial intelligence * Boosting
