Summary of Dependence Induced Representations, by Xiangxiang Xu et al.
Dependence Induced Representations
by Xiangxiang Xu, Lizhong Zheng
First submitted to arxiv on: 22 Nov 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 The paper investigates the problem of learning feature representations from two random variables that are dependent. It provides sufficient and necessary conditions for these dependence-induced representations, showing connections to Hirschfeld-Gebelein-Rényi (HGR) maximal correlation functions and minimal sufficient statistics. The authors characterize a large family of loss functions that can learn these representations, including cross entropy, hinge loss, and their regularized variants. This family allows features learned from different losses to be expressed as the composition of a loss-dependent function and the maximal correlation function, revealing a key connection between representations learned from different losses. Additionally, the paper gives a statistical interpretation of the neural collapse phenomenon observed in deep classifiers. The learning design based on feature separation is also presented, enabling hyperparameter tuning during inference. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper explores how to learn important features from two connected variables. It figures out what makes these features special and shows they’re related to other important concepts in statistics. The researchers then find a big group of ways to measure the success of learning these features, including some common methods used in artificial intelligence like cross entropy and hinge loss. This helps us understand how different ways of measuring success can lead to similar or different results. The paper also explains why some AI models might not work well with certain types of data, which is helpful for building better machines. |
Keywords
» Artificial intelligence » Cross entropy » Hinge loss » Hyperparameter » Inference
