Summary of Improving Generalization Of Deep Neural Networks by Optimum Shifting, By Yuyan Zhou et al.
Improving Generalization of Deep Neural Networks by Optimum Shifting
by Yuyan Zhou, Ye Li, Lei Feng, Sheng-Jun Huang
First submitted to arxiv on: 23 May 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 The proposed method, optimum shifting, modifies a neural network’s parameters from a sharp minimum to a flatter one while maintaining the same training loss value. This is achieved by treating matrix multiplications as under-determined linear equations and adjusting parameters through constrained optimization. The technique leverages Neural Collapse theory to reduce computational costs and increase degrees of freedom. Experimental results across various deep architectures on benchmark datasets demonstrate the effectiveness of optimum shifting in improving generalization ability. |
Low | GrooveSquid.com (original content) | Low Difficulty Summary Optimum shifting is a new way to help neural networks learn better. It takes an existing network and changes its parameters to make it more robust, while keeping the same performance during training. This is done by treating parts of the network as math problems that can be solved. The technique uses a concept called Neural Collapse to make it faster and more flexible. Tests on different types of networks and datasets show that optimum shifting improves generalization. |
Keywords
» Artificial intelligence » Generalization » Neural network » Optimization