Picture for Berfin Simsek

Berfin Simsek

Learning Gaussian Multi-Index Models with Gradient Flow: Time Complexity and Directional Convergence

Add code
Nov 13, 2024
Viaarxiv icon

Learning Associative Memories with Gradient Descent

Add code
Feb 28, 2024
Viaarxiv icon

The Loss Landscape of Shallow ReLU-like Neural Networks: Stationary Points, Saddle Escaping, and Network Embedding

Add code
Feb 08, 2024
Viaarxiv icon

Expand-and-Cluster: Exact Parameter Recovery of Neural Networks

Add code
Apr 25, 2023
Viaarxiv icon

Understanding out-of-distribution accuracies through quantifying difficulty of test samples

Add code
Mar 28, 2022
Figure 1 for Understanding out-of-distribution accuracies through quantifying difficulty of test samples
Figure 2 for Understanding out-of-distribution accuracies through quantifying difficulty of test samples
Figure 3 for Understanding out-of-distribution accuracies through quantifying difficulty of test samples
Figure 4 for Understanding out-of-distribution accuracies through quantifying difficulty of test samples
Viaarxiv icon

Weight-space symmetry in deep networks gives rise to permutation saddles, connected by equal-loss valleys across the loss landscape

Add code
Jul 05, 2019
Figure 1 for Weight-space symmetry in deep networks gives rise to permutation saddles, connected by equal-loss valleys across the loss landscape
Figure 2 for Weight-space symmetry in deep networks gives rise to permutation saddles, connected by equal-loss valleys across the loss landscape
Figure 3 for Weight-space symmetry in deep networks gives rise to permutation saddles, connected by equal-loss valleys across the loss landscape
Figure 4 for Weight-space symmetry in deep networks gives rise to permutation saddles, connected by equal-loss valleys across the loss landscape
Viaarxiv icon