Tensor Decomposition Methods in Neural Networks for Frugality

Explore a 24-minute conference talk from the Erwin Schrödinger International Institute's Thematic Programme on "Infinite-dimensional Geometry: Theory and Applications" that delves into tensor decomposition methods and their application in neural networks. Learn how these mathematical techniques can be leveraged both for compressing existing neural networks and for growing networks during the training process, with a focus on achieving frugality in neural network architectures. Discover the intersection of advanced mathematical concepts and practical machine learning applications as presented at this prestigious mathematical physics institute.