Summary of Flavors Of Margin: Implicit Bias Of Steepest Descent in Homogeneous Neural Networks, by Nikolaos Tsilivis et al.
Flavors of Margin: Implicit Bias of Steepest Descent in Homogeneous Neural Networks
by Nikolaos Tsilivis, Gal Vardi, Julia Kempe
First submitted to arxiv on: 29 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: 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 This paper investigates the implicit bias of various steepest descent algorithms, including gradient descent, sign descent, and coordinate descent, in deep homogeneous neural networks. The authors prove that an algorithm-dependent geometric margin increases once the networks reach perfect training accuracy and characterize the late-stage bias of these algorithms. They also define a generalized notion of stationarity for optimization problems and show how these algorithms reduce a Bregman divergence, which measures proximity to stationary points of a margin-maximization problem. Experimentally, the authors explore the trajectories of neural networks optimized with different steepest descent algorithms, revealing connections to the implicit bias of Adam. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper looks at how some popular AI training methods can be biased in certain ways. These methods include things like gradient descent and sign descent, which are used to train deep learning models called neural networks. The authors show that once these models become really good at fitting the data they’re trained on, they start to develop biases based on the method used to train them. They also explore how these biases can affect the performance of the models. |
Keywords
» Artificial intelligence » Deep learning » Gradient descent » Optimization
