Convergence of Gradient Descent with Small Initialization for Unregularized Matrix Completionby Jianhao Ma, Salar FattahiFirst…
On Differentially Private Subspace Estimation in a Distribution-Free Settingby Eliad TsfadiaFirst submitted to arxiv on:…
Optimizing Predictive AI in Physical Design Flows with Mini Pixel Batch Gradient Descentby Haoyu Yang,…
Loss Landscape of Shallow ReLU-like Neural Networks: Stationary Points, Saddle Escape, and Network Embeddingby Frank…
Implicit Bias and Fast Convergence Rates for Self-attentionby Bhavya Vasudeva, Puneesh Deora, Christos ThrampoulidisFirst submitted…
Graph Cuts with Arbitrary Size Constraints Through Optimal Transportby Chakib Fettal, Lazhar Labiod, Mohamed NadifFirst…
Decentralized Sporadic Federated Learning: A Unified Algorithmic Framework with Convergence Guaranteesby Shahryar Zehtabi, Dong-Jun Han,…
The Benefits of Reusing Batches for Gradient Descent in Two-Layer Networks: Breaking the Curse of…
Data-induced multiscale losses and efficient multirate gradient descent schemesby Juncai He, Liangchen Liu, Yen-Hsi Richard…
Discounted Adaptive Online Learning: Towards Better Regularizationby Zhiyu Zhang, David Bombara, Heng YangFirst submitted to…