Summary of Emergent Equivariance in Deep Ensembles, by Jan E. Gerken and Pan Kessel
Emergent Equivariance in Deep Ensembles
by Jan E. Gerken, Pan Kessel
First submitted to arxiv on: 5 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 This research paper presents a groundbreaking discovery in deep learning, showing that deep ensembles become equivariant for all inputs and training times by incorporating data augmentation. This breakthrough holds true even when considering off-manifold predictions and is applicable to any architecture in the infinite width limit. The study’s findings are attributed to the emergent property of collective prediction from individual ensemble members, which itself is not equivariant. To derive this result, the researchers leveraged neural tangent kernel theory, subsequently verifying their insights through comprehensive numerical experiments. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary In simple terms, this paper reveals that a special type of machine learning model called deep ensembles becomes better at understanding and representing patterns in data when it’s given extra practice with fake versions of that data. This improvement happens even when the model is making predictions about things that aren’t actually in the training data. The research team used mathematical theories to understand why this works, and then tested their ideas using lots of computer simulations. |
Keywords
* Artificial intelligence * Data augmentation * Deep learning * Machine learning
