Abstract
We consider a distributed multi-agent network system where the goal is to minimize an objective function that can be written as the sum of component functions, each of which is known (with stochastic errors) to a specific network agent. We propose an asynchronous algorithm that is motivated by a random gossip scheme where each agent has a local Poisson clock. At each tick of its local clock, the agent averages its estimate with a randomly chosen neighbor and adjusts the average using the gradient of its local function that is computed with stochastic errors. We investigate the convergence properties of the algorithm for two different classes of functions: differentiable but not necessarily convex and convex but not necessarily differentiable.
| Original language | English (US) |
|---|---|
| Title of host publication | Proceedings of the 2009 International Conference on Game Theory for Networks, GameNets '09 |
| Pages | 80-81 |
| Number of pages | 2 |
| DOIs | |
| State | Published - 2009 |
| Externally published | Yes |
| Event | 2009 International Conference on Game Theory for Networks, GameNets '09 - Istanbul, Turkey Duration: May 13 2009 → May 15 2009 |
Other
| Other | 2009 International Conference on Game Theory for Networks, GameNets '09 |
|---|---|
| Country | Turkey |
| City | Istanbul |
| Period | 5/13/09 → 5/15/09 |
Fingerprint
ASJC Scopus subject areas
- Computer Networks and Communications
- Computer Vision and Pattern Recognition
Cite this
Asynchronous gossip algorithms for stochastic optimization. / Sundhar Ram, S.; Nedich, Angelia; Veeravalli, V. V.
Proceedings of the 2009 International Conference on Game Theory for Networks, GameNets '09. 2009. p. 80-81 5137386.Research output: Chapter in Book/Report/Conference proceeding › Conference contribution
}
TY - GEN
T1 - Asynchronous gossip algorithms for stochastic optimization
AU - Sundhar Ram, S.
AU - Nedich, Angelia
AU - Veeravalli, V. V.
PY - 2009
Y1 - 2009
N2 - We consider a distributed multi-agent network system where the goal is to minimize an objective function that can be written as the sum of component functions, each of which is known (with stochastic errors) to a specific network agent. We propose an asynchronous algorithm that is motivated by a random gossip scheme where each agent has a local Poisson clock. At each tick of its local clock, the agent averages its estimate with a randomly chosen neighbor and adjusts the average using the gradient of its local function that is computed with stochastic errors. We investigate the convergence properties of the algorithm for two different classes of functions: differentiable but not necessarily convex and convex but not necessarily differentiable.
AB - We consider a distributed multi-agent network system where the goal is to minimize an objective function that can be written as the sum of component functions, each of which is known (with stochastic errors) to a specific network agent. We propose an asynchronous algorithm that is motivated by a random gossip scheme where each agent has a local Poisson clock. At each tick of its local clock, the agent averages its estimate with a randomly chosen neighbor and adjusts the average using the gradient of its local function that is computed with stochastic errors. We investigate the convergence properties of the algorithm for two different classes of functions: differentiable but not necessarily convex and convex but not necessarily differentiable.
UR - http://www.scopus.com/inward/record.url?scp=70349977432&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=70349977432&partnerID=8YFLogxK
U2 - 10.1109/GAMENETS.2009.5137386
DO - 10.1109/GAMENETS.2009.5137386
M3 - Conference contribution
AN - SCOPUS:70349977432
SN - 9781424441778
SP - 80
EP - 81
BT - Proceedings of the 2009 International Conference on Game Theory for Networks, GameNets '09
ER -