Publications

Browse publications by type, listed in reverse chronological order. * Denotes equal contribution.

Journal Articles 4 publications

[J1] Sharp High-Probability Rates for Nonlinear SGD under Heavy-Tailed Noise via Symmetrization 2026

A. Armacki, D. Bajović, D. Jakovetić, S. Kar.

IEEE Transactions on Information Theory, 2026.

[J2] Large Deviation Upper Bounds and Improved MSE Rates of Nonlinear SGD: Heavy-Tailed Noise and Power of Symmetry 2026

A. Armacki*, S. Yu*, D. Bajović, D. Jakovetić, S. Kar.

SIAM Journal on Optimization, 2026.

Published version ↗ Preprint ↗

[J3] Distributed Center-Based Clustering: A Unified Framework 2025

A. Armacki, D. Bajović, D. Jakovetić, S. Kar.

IEEE Transactions on Signal Processing, 2025.

Published version ↗ Preprint ↗

[J4] A One-shot Framework for Distributed Clustered Learning in Heterogeneous Environments 2024

A. Armacki, D. Bajović, D. Jakovetić, S. Kar.

IEEE Transactions on Signal Processing, 2024.

Published version ↗ Preprint ↗

Conference Papers 9 publications

[C1] Tight Long-Term Tail Decay of (Clipped) SGD in Non-Convex Optimization 2026

A. Armacki, D. Bajović, D. Jakovetić, S. Kar, A. H. Sayed.

39th Conference on Learning Theory (COLT), 2026.

Published version ↗ Preprint ↗

[C2] High-Probability Convergence Guarantees of Decentralized SGD 2026

A. Armacki, A. H. Sayed.

43rd International Conference on Machine Learning (ICML), 2026.

Published version ↗ Preprint ↗

[C3] Toward Understanding the Improved Robustness to Initialization in Distributed Clustering 2025

A. Armacki, S. Kar.

33rd European Signal Processing Conference (EUSIPCO), 2025.

Published version ↗ Preprint ↗

[C4] High-probability Convergence Bounds for Online Nonlinear Stochastic Gradient Descent under Heavy-tailed Noise 2025

A. Armacki, S. Yu, P. Sharma, G. Joshi, D. Bajović, D. Jakovetić, S. Kar.

28th International Conference on Artificial Intelligence and Statistics (AISTATS), 2025.

Published version ↗ Preprint ↗

[C5] Distributed Gradient Clustering: Convergence and the Effect of Initialization 2024

A. Armacki*, H. Sharma*, D. Bajović, D. Jakovetić, M. Chakraborty, S. Kar.

58th Asilomar Conference on Signals, Systems, and Computers, 2024.

Published version ↗ Preprint ↗

[C6] Communication Efficient Model-aware Federated Learning for Visual Crowd Counting and Density Estimation in Smart Cities 2023

A. Armacki, N. Milošević, D. Bajović, S. Kar, D. Jakovetić, A. Bakhtiarnia, L. Esterle, A. Muscat, T. Festi.

31st European Signal Processing Conference (EUSIPCO), 2023.

Published version ↗ Preprint ↗

[C7] Personalized Federated Learning via Convex Clustering 2022

A. Armacki, D. Bajović, D. Jakovetić, S. Kar.

IEEE International Smart Cities Conference (ISC2), 2022.

Published version ↗ Preprint ↗

[C8] Gradient Based Clustering 2022

A. Armacki, D. Bajović, D. Jakovetić, S. Kar.

39th International Conference on Machine Learning (ICML), 2022.

Published version ↗ Preprint ↗

[C9] Distributed Trust-Region Method With First Order Models 2019

A. Armacki, D. Jakovetić, N. Krejić, N. Krklec Jerinkić.

IEEE EUROCON 18th International Conference on Smart Technologies, 2019.

Published version ↗ Preprint ↗

Books and Textbooks 1 publication

[B1] Distributed Optimization with Applications 2020

D. Jakovetić, A. Armacki.

Graduate textbook, Faculty of Sciences, University of Novi Sad, 2020.

Textbook (PDF) ↗

Preprints 3 publications

[P1] Decentralized SGD under Heavy-Tailed Noise: Optimal Convergence Rates and the Role of Gradient Clipping 2026

A. Armacki, H. Cai, A. H. Sayed.

Under review, 2026.

Preprint ↗

[P2] A Unified Zeroth-Order Approach for Decentralized Minimax Optimization 2026

H. Cai, Y. Zhao, A. Armacki, J. Chen, A. H. Sayed.

arXiv preprint, 2026.

Preprint ↗

[P3] High-Probability Convergence in Decentralized Stochastic Optimization with Gradient Tracking 2026

A. Armacki, H. Cai, A. H. Sayed.

Under review, 2026.

Preprint ↗

Aleksandar Armacki

Publications