Button to scroll to the top of the page.


Monthly View
By Month
Weekly View
By Week
Daily View
Download as iCal file
Angelia Nedich, RLM 5.104: Distributed Algorithms for Optimization over Graphs
Monday, March 04, 2019, 04:00pm - 05:00pm
We will discuss the problems of distributed optimization overtime-varying graphs. The focus will be on more recent effortsto design such methods that can match the performance of theircentralized counterparts. These fast algorithms will be presentedfor both directed and undirected time-varying graphs. These algorithmscombine the traditional gradient-based methods with tracking-mechanismsthat allow for learning the missing information from the localneighbors in the graph. We will discuss the convergence and convergencerate results of these methods, where the convergence rate capturesthe problem related quantities (such as strong convexity constantand gradient Lipschitz constant) and the properties of the graphs,such as the eigenvalues of the associated weighted Laplaciansor mixing weights. We will also consider scalability of thesealgorithms in terms of the number of nodes (agents) in the graph.Then, we will present more specialized algorithms for optimizationover static graphs whose performance matches the performanceof their centralized counterparts (up to a logarithmic factor).
Location: RLM 5.104

Math Calendar Login