Picture for Konstantin Burlachenko

Konstantin Burlachenko

BurTorch: Revisiting Training from First Principles by Coupling Autodiff, Math Optimization, and Systems

Add code
Mar 18, 2025
Viaarxiv icon

Unlocking FedNL: Self-Contained Compute-Optimized Implementation

Add code
Oct 11, 2024
Figure 1 for Unlocking FedNL: Self-Contained Compute-Optimized Implementation
Figure 2 for Unlocking FedNL: Self-Contained Compute-Optimized Implementation
Figure 3 for Unlocking FedNL: Self-Contained Compute-Optimized Implementation
Figure 4 for Unlocking FedNL: Self-Contained Compute-Optimized Implementation
Viaarxiv icon

PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression

Add code
May 23, 2024
Viaarxiv icon

Error Feedback Reloaded: From Quadratic to Arithmetic Mean of Smoothness Constants

Add code
Feb 16, 2024
Viaarxiv icon

Federated Learning is Better with Non-Homomorphic Encryption

Add code
Dec 04, 2023
Viaarxiv icon

Error Feedback Shines when Features are Rare

Add code
May 24, 2023
Viaarxiv icon

Federated Learning with Regularized Client Participation

Add code
Feb 28, 2023
Viaarxiv icon

Personalized Federated Learning with Communication Compression

Add code
Sep 12, 2022
Figure 1 for Personalized Federated Learning with Communication Compression
Figure 2 for Personalized Federated Learning with Communication Compression
Figure 3 for Personalized Federated Learning with Communication Compression
Figure 4 for Personalized Federated Learning with Communication Compression
Viaarxiv icon

Federated Optimization Algorithms with Random Reshuffling and Gradient Compression

Add code
Jun 14, 2022
Figure 1 for Federated Optimization Algorithms with Random Reshuffling and Gradient Compression
Figure 2 for Federated Optimization Algorithms with Random Reshuffling and Gradient Compression
Figure 3 for Federated Optimization Algorithms with Random Reshuffling and Gradient Compression
Figure 4 for Federated Optimization Algorithms with Random Reshuffling and Gradient Compression
Viaarxiv icon

Sharper Rates and Flexible Framework for Nonconvex SGD with Client and Data Sampling

Add code
Jun 05, 2022
Figure 1 for Sharper Rates and Flexible Framework for Nonconvex SGD with Client and Data Sampling
Figure 2 for Sharper Rates and Flexible Framework for Nonconvex SGD with Client and Data Sampling
Figure 3 for Sharper Rates and Flexible Framework for Nonconvex SGD with Client and Data Sampling
Figure 4 for Sharper Rates and Flexible Framework for Nonconvex SGD with Client and Data Sampling
Viaarxiv icon