Summary of On the Hardness Of Learning Under Symmetries, by Bobak T. Kiani et al.
On the hardness of learning under symmetries
by Bobak T. Kiani, Thien Le, Hannah Lawrence, Stefanie Jegelka, Melanie Weber
First submitted to arxiv on: 3 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Data Structures and Algorithms (cs.DS); Statistics Theory (math.ST); Machine Learning (stat.ML)
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 abstract explores the limitations of learning equivariant neural networks using gradient descent, highlighting that even with known symmetries, the complexity of learning shallow networks can be exponential. The researchers investigate whether symmetries are sufficient to alleviate the hardness of learning and find that lower bounds for various network types, including graph and convolutional networks, scale either superpolynomially or exponentially in the input dimension. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study reveals that even when incorporating known symmetries into neural networks, they can still be difficult to learn using gradient descent. The researchers demonstrate this by showing lower bounds for different types of neural networks, including those with graph and convolutional structures. This means that just because a network has symmetries, it doesn’t necessarily make learning it easier. |
Keywords
* Artificial intelligence * Gradient descent
