arXiv subjects

Sujay Sanghavi

Publications and source records attributed to Sujay Sanghavi.

114 records · Page 7Linked to original sources

arXiv2008

Message-passing for Maximum Weight Independent Set

We investigate the use of message-passing algorithms for the problem of finding the max-weight independent set (MWIS) in a graph. First, we study the performance of the classical loopy max-product belief propagation. We show that each fixed point estimate of max-product can be mapped in a natural way to an extreme point of the LP polytope associated with the MWIS problem. However, this extreme point may not be the one that maximizes the value of node weights; the particular extreme point at final convergence depends on the initialization of max-product. We then show that if max-product is started from the natural initialization of uninformative messages, it always solves the correct LP -- if it converges. This result is obtained via a direct analysis of the iterative algorithm, and cannot be obtained by looking only at fixed points. The tightness of the LP relaxation is thus necessary for max-product optimality, but it is not sufficient. Motivated by this observation, we show that a simple modification of max-product becomes gradient descent on (a convexified version of) the dual of the LP, and converges to the dual optimum. We also develop a message-passing algorithm that recovers the primal MWIS solution from the output of the descent algorithm. We show that the MWIS estimate obtained using these two algorithms in conjunction is correct when the graph is bipartite and the MWIS is unique. Finally, we show that any problem of MAP estimation for probability distributions over finite domains can be reduced to an MWIS problem. We believe this reduction will yield new insights and algorithms for MAP estimation.

cs.AI↗

arXiv2008

Tightness of LP via Max-product Belief Propagation

We investigate the question of tightness of linear programming (LP) relaxation for finding a maximum weight independent set (MWIS) in sparse random weighted graphs. We show that an edge-based LP relaxation is asymptotically tight for Erdos-Renyi graph $G(n,c/n)$ for $c \leq 2e$ and random regular graph $G(n,r)$ for $r\leq 4$ when node weights are i.i.d. with exponential distribution of mean 1. We establish these results, through a precise relation between the tightness of LP relaxation and convergence of the max-product belief propagation algorithm. We believe that this novel method of understanding structural properties of combinatorial problems through properties of iterative procedure such as the max-product should be of interest in its own right.

cs.DS↗

arXiv2007

Equivalence of LP Relaxation and Max-Product for Weighted Matching in General Graphs

Max-product belief propagation is a local, iterative algorithm to find the mode/MAP estimate of a probability distribution. While it has been successfully employed in a wide variety of applications, there are relatively few theoretical guarantees of convergence and correctness for general loopy graphs that may have many short cycles. Of these, even fewer provide exact ``necessary and sufficient'' characterizations. In this paper we investigate the problem of using max-product to find the maximum weight matching in an arbitrary graph with edge weights. This is done by first constructing a probability distribution whose mode corresponds to the optimal matching, and then running max-product. Weighted matching can also be posed as an integer program, for which there is an LP relaxation. This relaxation is not always tight. In this paper we show that \begin{enumerate} \item If the LP relaxation is tight, then max-product always converges, and that too to the correct answer. \item If the LP relaxation is loose, then max-product does not converge. \end{enumerate} This provides an exact, data-dependent characterization of max-product performance, and a precise connection to LP relaxation, which is a well-studied optimization technique. Also, since LP relaxation is known to be tight for bipartite graphs, our results generalize other recent results on using max-product to find weighted matchings in bipartite graphs.

cs.IT↗

arXiv2006

Gossiping with Multiple Messages

This paper investigates the dissemination of multiple pieces of information in large networks where users contact each other in a random uncoordinated manner, and users upload one piece per unit time. The underlying motivation is the design and analysis of piece selection protocols for peer-to-peer networks which disseminate files by dividing them into pieces. We first investigate one-sided protocols, where piece selection is based on the states of either the transmitter or the receiver. We show that any such protocol relying only on pushes, or alternatively only on pulls, is inefficient in disseminating all pieces to all users. We propose a hybrid one-sided piece selection protocol -- INTERLEAVE -- and show that by using both pushes and pulls it disseminates $k$ pieces from a single source to $n$ users in $10(k+\log n)$ time, while obeying the constraint that each user can upload at most one piece in one unit of time, with high probability for large $n$. An optimal, unrealistic centralized protocol would take $k+\log_2 n$ time in this setting. Moreover, efficient dissemination is also possible if the source implements forward erasure coding, and users push the latest-released coded pieces (but do not pull). We also investigate two-sided protocols where piece selection is based on the states of both the transmitter and the receiver. We show that it is possible to disseminate $n$ pieces to $n$ users in $n+O(\log n)$ time, starting from an initial state where each user has a unique piece.

cs.NI↗

arXiv2006

Intermediate Performance of Rateless Codes

Rateless/fountain codes are designed so that all input symbols can be recovered from a slightly larger number of coded symbols, with high probability using an iterative decoder. In this paper we investigate the number of input symbols that can be recovered by the same decoder, but when the number of coded symbols available is less than the total number of input symbols. Of course recovery of all inputs is not possible, and the fraction that can be recovered will depend on the output degree distribution of the code. In this paper we (a) outer bound the fraction of inputs that can be recovered for any output degree distribution of the code, and (b) design degree distributions which meet/perform close to this bound. Our results are of interest for real-time systems using rateless codes, and for Raptor-type two-stage designs.

cs.IT↗

arXiv2006

Distributed Link Scheduling with Constant Overhead

This paper proposes a new class of simple, distributed algorithms for scheduling in wireless networks. The algorithms generate new schedules in a distributed manner via simple local changes to existing schedules. The class is parameterized by integers $k\geq 1$. We show that algorithm $k$ of our class achieves $k/(k+2)$ of the capacity region, for every $k\geq 1$. The algorithms have small and constant worst-case overheads: in particular, algorithm $k$ generates a new schedule using {\em (a)} time less than $4k+2$ round-trip times between neighboring nodes in the network, and {\em (b)} at most three control transmissions by any given node, for any $k$. The control signals are explicitly specified, and face the same interference effects as normal data transmissions. Our class of distributed wireless scheduling algorithms are the first ones guaranteed to achieve any fixed fraction of the capacity region while using small and constant overheads that do not scale with network size. The parameter $k$ explicitly captures the tradeoff between control overhead and scheduler throughput performance and provides a tuning knob protocol designers can use to harness this trade-off in practice.

cs.NI↗