Summary of Semi-supervised One-shot Imitation Learning, by Philipp Wu and Kourosh Hakhamaneshi and Yuqing Du and Igor Mordatch and Aravind Rajeswaran and Pieter Abbeel
Semi-Supervised One-Shot Imitation Learning
by Philipp Wu, Kourosh Hakhamaneshi, Yuqing Du, Igor Mordatch, Aravind Rajeswaran, Pieter Abbeel
First submitted to arxiv on: 9 Aug 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 The paper introduces One-shot Imitation Learning (OSIL), which enables AI agents to learn from a single demonstration. Typically, OSIL requires numerous paired expert demonstrations, but this can be impractical. To overcome this limitation, the authors propose a semi-supervised OSIL setting, where the agent receives an unpaired dataset of trajectories and a small paired dataset with task labels. This setting mimics few-shot learning, requiring the agent to leverage weak supervision from the large dataset. The authors develop an algorithm for this setting, first learning an embedding space that clusters tasks uniquely. They then use this space to self-generate pairings between unpaired trajectories, allowing OSIL models to be trained with competitive performance using only weak supervision. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary OSIL is a way to teach AI agents new skills by showing them one example. Usually, you need many examples to train an AI agent, but that’s not always possible. In this paper, the authors solve this problem by giving the agent a big collection of actions (trajectories) with no labels and a small group of labeled examples. This helps the agent learn more efficiently and accurately. |
Keywords
» Artificial intelligence » Embedding space » Few shot » One shot » Semi supervised
