On the impact of activation and normalization in obtaining isometric embeddings at initialization

A Joudaki, H Daneshmand… - Advances in Neural …, 2023 - proceedings.neurips.cc
In this paper, we explore the structure of the penultimate Gram matrix in deep neural
networks, which contains the pairwise inner products of outputs corresponding to a batch of …

Wasserstein gradient flows of MMD functionals with distance kernel and Cauchy problems on quantile functions

R Duong, V Stein, R Beinert, J Hertrich… - arxiv preprint arxiv …, 2024 - arxiv.org
We give a comprehensive description of Wasserstein gradient flows of maximum mean
discrepancy (MMD) functionals $\mathcal F_\nu:=\text {MMD} _K^ 2 (\cdot,\nu) $ towards …

Position: -Algebraic Machine Learning Moving in a New Direction

Y Hashimoto, M Ikeda, H Kadri - Forty-first International Conference …, 2024 - openreview.net
Machine learning has a long collaborative tradition with several fields of mathematics, such
as statistics, probability and linear algebra. We propose a new direction for machine …

Analyzing the Geometric Structure of Deep Learning Decision Boundaries

M Geyer - 2023 - search.proquest.com
Training deep learning models is an incredibly effective method for finding function
approximators. However, understanding the behavior of these trained models from a first …