arXiv subjects

Akash Rakheja

Publications and source records attributed to Akash Rakheja.

4 recordsLinked to original sources

arXiv2021

Prophet Matching Meets Probing with Commitment

We consider the online stochastic matching problem for bipartite graphs where edges adjacent to an online node must be probed to determine if they exist, based on known edge probabilities. Our algorithms respect commitment, in that if a probed edge exists, it must be used in the matching. We study this matching problem subject to a downward-closed constraint on each online node's allowable edge probes. Our setting generalizes the commonly studied patience (or time-out) constraint which limits the number of probes that can be made to an online node's adjacent edges. We introduce a new LP that we prove is a relaxation of an optimal offline probing algorithm (the adaptive benchmark) and which overcomes the limitations of previous LP relaxations. (1) A tight $\frac{1}{2}$ ratio when the stochastic graph is generated from a known stochastic type graph where the $t^{th}$ online node is drawn independently from a known distribution $\scr{D}_{π(t)}$ and $π$ is chosen adversarially. We refer to this setting as the known i.d. stochastic matching problem with adversarial arrivals. (2) A $1-1/e$ ratio when the stochastic graph is generated from a known stochastic type graph where the $t^{th}$ online node is drawn independently from a known distribution $\scr{D}_{π(t)}$ and $π$ is a random permutation. We refer to this setting as the known i.d. stochastic matching problem with random order arrivals. Our results improve upon the previous best competitive ratio of $0.46$ in the known i.i.d. setting against the standard adaptive benchmark. Moreover, we are the first to study the prophet secretary matching problem in the context of probing, where we match the best known classical result.

cs.DM↗

arXiv2021

Greedy Approaches to Online Stochastic Matching

Within the context of stochastic probing with commitment, we consider the online stochastic matching problem; that is, the one-sided online bipartite matching problem where edges adjacent to an online node must be probed to determine if they exist based on edge probabilities that become known when an online vertex arrives. If a probed edge exists, it must be used in the matching (if possible). We consider the competitiveness of online algorithms in both the adversarial order model (AOM) and the random order model (ROM). More specifically, we consider a bipartite stochastic graph $G = (U,V,E)$ where $U$ is the set of offline vertices, $V$ is the set of online vertices and $G$ has edge probabilities $(p_{e})_{e \in E}$ and edge weights $(w_{e})_{e \in E}$. Additionally, $G$ has probing constraints $(\scr{C}_{v})_{v \in V}$, where $\scr{C}_v$ indicates which sequences of edges adjacent to an online vertex $v$ can be probed. We assume that $U$ is known in advance, and that $\scr{C}_v$, together with the edge probabilities and weights adjacent to an online vertex are only revealed when the online vertex arrives. This model generalizes the various settings of the classical bipartite matching problem, and so our main contribution is in making progress towards understanding which classical results extend to the stochastic probing model.

cs.DS↗

arXiv2021

Bipartite Stochastic Matching: Online, Random Order, and I.I.D. Models

Within the context of stochastic probing with commitment, we consider the online stochastic matching problem; that is, the one sided online bipartite matching problem where edges adjacent to an online node must be probed to determine if they exist, based on known edge probabilities. If a probed edge exists, it must be used in the matching (if possible). We study this problem in the generality of a patience (or budget) constraint which limits the number of probes that can be made to edges adjacent to an online node. The patience constraint results in modelling and computational efficiency issues that are not encountered in the special cases of unit patience and full (i.e., unlimited) patience. The stochastic matching problem leads to a variety of settings. Our main contribution is to provide a new LP relaxation and a unified approach for establishing new and improved competitive bounds in three different input model settings (namely, adversarial, random order, and known i.i.d.). In all these settings, the algorithm does not have any control on the ordering of the online nodes. We establish competitive bounds in these settings, all of which generalize the standard non-stochastic setting when edges do not need to be probed (i.e., exist with certainty). All of our competitive ratios hold for arbitrary edge probabilities and patience constraints: (1) A $1-1/e$ ratio when the stochastic graph is known, offline vertices are weighted and online arrivals are adversarial. (2) A $1-1/e$ ratio when the stochastic graph is known, edges are weighted, and online arrivals are given in random order (i.e., in ROM, the random order model). (3) A $1-1/e$ ratio when online arrivals are drawn i.i.d. from a known stochastic type graph and edges are weighted. (4) A (tight) $1/e$ ratio when the stochastic graph is unknown, edges are weighted and online arrivals are given in random order.

cs.DS↗

arXiv2019

Electronic markets with multiple submodular buyers

We discuss the problem of setting prices in an electronic market that has more than one buyer. We assume that there are self-interested sellers each selling a distinct item that has an associated cost. Each buyer has a submodular valuation for purchasing any subset of items. The goal of the sellers is to set a price for their item such that their profit from possibly selling their item to the buyers is maximized. Our most comprehensive results concern a multi copy setting where each seller has m copies of their item and there are m buyers. In this setting, we give a necessary and sufficient condition for the existence of market clearing pure Nash equilibrium. We also show that not all equilibria are market clearing even when this condition is satisfied contrary to what was shown in the case of a single buyer in [2]. Finally, we investigate the pricing problem for multiple buyers in the limited supply setting when each seller only has a single copy of their item.

cs.GT↗