Молимо вас користите овај идентификатор за цитирање или овај линк до ове ставке:
https://open.uns.ac.rs/handle/123456789/18249
Назив: | Convergence rates of distributed nesterov-like gradient methods on random networks | Аутори: | Jakovetić Dušan Joao Xavier Jose Moura |
Датум издавања: | 2014 | Часопис: | IEEE Transactions on Signal Processing | Сажетак: | We consider distributed optimization in random networks where N nodes cooperatively minimize the sum \sum -{i=1}^{N} f-{i}(x) of their individual convex costs. Existing literature proposes distributed gradient-like methods that are computationally cheap and resilient to link failures, but have slow convergence rates. In this paper, we propose accelerated distributed gradient methods that 1) are resilient to link failures; 2) computationally cheap; and 3) improve convergence rates over other gradient methods. We model the network by a sequence of independent, identically distributed random matrices \{W(k)\} drawn from the set of symmetric, stochastic matrices with positive diagonals. The network is connected on average and the cost functions are convex, differentiable, with Lipschitz continuous and bounded gradients. We design two distributed Nesterov-like gradient methods that modify the D-NG and D-NC methods that we proposed for static networks. We prove their convergence rates in terms of the expected optimality gap at the cost function. Let k and {\cal K} be the number of per-node gradient evaluations and per-node communications, respectively. Then the modified D-NG achieves rates O(\log k/k) and O(\log {\cal K}/ {\cal K}) , and the modified D-NC rates O(1/k^{2}) and O(1/ {\cal K}^{2-\xi }), where \xi >0 is arbitrarily small. For comparison, the standard distributed gradient method cannot do better than \Omega (1/k^{2/3}) and \Omega (1/ {\cal K}^{2/3}), on the same class of cost functions (even for static networks). Simulation examples illustrate our analytical findings. © 1991-2012 IEEE. | URI: | https://open.uns.ac.rs/handle/123456789/18249 | ISSN: | 1053-587X | DOI: | 10.1109/TSP.2013.2291221 |
Налази се у колекцијама: | PMF Publikacije/Publications |
Приказати целокупан запис ставки
Преглед/и станица
22
Протекла недеља
3
3
Протекли месец
0
0
проверено 10.05.2024.
Google ScholarTM
Проверите
Алт метрика
Ставке на DSpace-у су заштићене ауторским правима, са свим правима задржаним, осим ако није другачије назначено.