random matrix theory
A branch of mathematics that deals with the properties of matrices that have random entries. In AI, it is used to analyze the behavior of high-dimensional data and neural networks, especially in understanding their performance and stability.
- Bayes optimal learning of attention-indexed models
- Non-Asymptotic Analysis Of Data Augmentation For Precision Matrix Estimation
- Optimal Spectral Transitions in High-Dimensional Multi-Index Models
- Small Singular Values Matter: A Random Matrix Analysis of Transformer Models
- Solving Neural Min-Max Games: The Role of Architecture, Initialization & Dynamics
- Spectral Analysis of Representational Similarity with Limited Neurons
- Spectral Estimation with Free Decompression