Neurips 2024 - Symmetries in Overparametrized Neural Networks: A Mean Field View
Conference Talk, NeurIPS Conference (online), Vancouver, Canada
As a spotlight paper at NeurIPS, we presented our work in a 5 min video which was available through the conference website, as well as a poster in one of the main poster sessions. The talk outlined the different elements of our paper (together with Joaquín Fontbona) on the effect of data-symmetries in neural-network training at the Mean Field level.
