Optimal Neural Network Compressors and the Manifold Hypothesis

Optimal Neural Network Compressors and the Manifold Hypothesis

Explore the intersection of information theory and machine learning in this 36-minute lecture by Aaron Wagner from Cornell University. Delve into the concept of optimal neural network compressors and their relationship to the manifold hypothesis. Gain insights into how these principles contribute to the development of trustworthy machine learning systems. Examine the theoretical foundations and practical implications of compressing neural networks while maintaining their performance and reliability.