← Search

Ankit Sonthalia

2 accepted papers

2025

Do Deep Neural Network Solutions Form a Star Domain?

ICLR 2025poster

It has recently been conjectured that neural network solution sets reachable via stochastic gradient descent (SGD) are convex, considering permutation invariances. This means that a linear path can connect two independent solutions with low loss, given the weights of one of the models are appropriat…