Samuel Horváth

MT-DAO: Multi-Timescale Distributed Adaptive Optimizers with Local Updates

DES-LOC: Desynced Low Communication Adaptive Optimizers for Foundation Models

LoFT: Low-Rank Adaptation That Behaves Like Full Fine-Tuning

Stochastic Self-Organization in Multi-Agent Systems

Double Momentum and Error Feedback for Clipping with Fast Rates and Differential Privacy

Loss Transformation Invariance of the Damped Newton Methods

Convergence of Clipped-SGD for Convex $(L_0,L_1)$-Smooth Optimization with Heavy-Tailed Noise

Top-K Structure Search with Solution Path

Initialization using Update Approximation is a Silver Bullet for Extremely Efficient Low-Rank Fine-Tuning

Preconditioned Norms: A Unified Framework for Steepest Descent, Quasi-Newton and Adaptive Methods

Selective Collaboration for Robust Federated Learning

Simple Stepsizes for Quasi-Newton Methods with Global Convergence Guarantees

Where Redundancy Lives: Stage-Aware Block Saliency in Skip-Connected Models

Revisiting One-Shot Pruning with Scalable Second-Order Approximations

Methods with Local Steps and Random Reshuffling for Generally Smooth Non-Convex Federated Optimization

Differentially Private Clipped-SGD: High-Probability Convergence with Arbitrary Clipping Level

Methods for Convex $(L_0,L_1)$-Smooth Optimization: Clipping, Acceleration, and Adaptivity

Aequa: Fair Model Rewards in Collaborative Learning via Slimmable Networks

Momentum and Error Feedback for Clipping with Fast Rates and Differential Privacy

Clipping Improves Adam and AdaGrad when the Noise Is Heavy-Tailed

FRUGAL: Memory-Efficient Optimization by Reducing State Overhead for Scalable Training

FRUGAL: Memory-Efficient Optimization by Reducing State Overhead for Scalable Training

Clipping Improves Adam-Norm and AdaGrad-Norm when the Noise Is Heavy-Tailed

Federated Learning Can Find Friends That Are Advantageous

Flashback: Understanding and Mitigating Forgetting in Federated Learning

Collaborative and Efficient Personalization with Mixtures of Adaptors

FedPeWS: Personalized Warmup via Subnetworks for Enhanced Heterogeneous Federated Learning