Picture for Grigory Malinovsky

Grigory Malinovsky

King Abdullah University of Science and Technology

MicroAdam: Accurate Adaptive Optimization with Low Space Overhead and Provable Convergence

Add code
May 24, 2024
Viaarxiv icon

Streamlining in the Riemannian Realm: Efficient Riemannian Optimization with Loopless Variance Reduction

Mar 11, 2024
Figure 1 for Streamlining in the Riemannian Realm: Efficient Riemannian Optimization with Loopless Variance Reduction
Viaarxiv icon

MAST: Model-Agnostic Sparsified Training

Add code
Nov 27, 2023
Figure 1 for MAST: Model-Agnostic Sparsified Training
Figure 2 for MAST: Model-Agnostic Sparsified Training
Figure 3 for MAST: Model-Agnostic Sparsified Training
Viaarxiv icon

Byzantine Robustness and Partial Participation Can Be Achieved Simultaneously: Just Clip Gradient Differences

Nov 23, 2023
Viaarxiv icon

Improving Accelerated Federated Learning with Compression and Importance Sampling

Jun 05, 2023
Figure 1 for Improving Accelerated Federated Learning with Compression and Importance Sampling
Viaarxiv icon

A Guide Through the Zoo of Biased SGD

Add code
May 25, 2023
Figure 1 for A Guide Through the Zoo of Biased SGD
Figure 2 for A Guide Through the Zoo of Biased SGD
Figure 3 for A Guide Through the Zoo of Biased SGD
Figure 4 for A Guide Through the Zoo of Biased SGD
Viaarxiv icon

Federated Learning with Regularized Client Participation

Feb 28, 2023
Figure 1 for Federated Learning with Regularized Client Participation
Figure 2 for Federated Learning with Regularized Client Participation
Figure 3 for Federated Learning with Regularized Client Participation
Figure 4 for Federated Learning with Regularized Client Participation
Viaarxiv icon

TAMUNA: Accelerated Federated Learning with Local Training and Partial Participation

Feb 20, 2023
Figure 1 for TAMUNA: Accelerated Federated Learning with Local Training and Partial Participation
Figure 2 for TAMUNA: Accelerated Federated Learning with Local Training and Partial Participation
Figure 3 for TAMUNA: Accelerated Federated Learning with Local Training and Partial Participation
Viaarxiv icon

Can 5th Generation Local Training Methods Support Client Sampling? Yes!

Jan 06, 2023
Figure 1 for Can 5th Generation Local Training Methods Support Client Sampling? Yes!
Figure 2 for Can 5th Generation Local Training Methods Support Client Sampling? Yes!
Figure 3 for Can 5th Generation Local Training Methods Support Client Sampling? Yes!
Figure 4 for Can 5th Generation Local Training Methods Support Client Sampling? Yes!
Viaarxiv icon

An Optimal Algorithm for Strongly Convex Min-min Optimization

Dec 29, 2022
Figure 1 for An Optimal Algorithm for Strongly Convex Min-min Optimization
Figure 2 for An Optimal Algorithm for Strongly Convex Min-min Optimization
Viaarxiv icon