Literature

Aggregators

• arxiv
• arxiv-sanity

Articles

Topology related:

• Primer on Persistent Homology
• Topological Data Analysis
  • from the 2020 ICML
  • from the 2019 ICML

On the Universal Approximation Theorem and computational power:

• Approximation by Superpositions of Sigmoidal Functions by G. Cybenko (1989)
• Multilayer Feedforward Networks With a Nonpolynomial Activation Function Can Approximate Any Function (1993)
• Approximation theory of the MLP model in neural networks (1999)
• On the Computational Power of Neural Nets (1992)
• Universality of deep convolutional neural networks (2018) arxiv
• Universal Function Approximation by Deep Neural Nets with Bounded Width and ReLU Activations (2019)
• A closer look at the approximation capabilities of neural networks

Datasets

• Text
  • The Pile

Textbooks

• ESLII

Transformers

• Nyströmformer: A Nyström-Based Algorithm for Approximating Self-Attention
• video explaining this paper

Manifold/Statistical Learning

• The mathematical foundations of Manifold Learning
• Deep Nets for Local Manifold Learning
• Geometric deep learning: going beyond Euclidean data