Pair-Matching: Link Prediction with Adaptive Queries - ENSAE Paris Access content directly
Journal Articles Mathematical Statistics and Learning Year : 2024

Pair-Matching: Link Prediction with Adaptive Queries

Abstract

The pair-matching problem appears in many applications where one wants to discover good matches between pairs of entities or individuals. Formally, the set of individuals is represented by the nodes of a graph where the edges, unobserved at first, represent the good matches. The algorithm queries pairs of nodes and observes the presence/absence of edges. Its goal is to discover as many edges as possible with a fixed budget of queries. Pair-matching is a particular instance of multi-armed bandit problem in which the arms are pairs of individuals and the rewards are edges linking these pairs. This bandit problem is non-standard though, as each arm can only be played once. Given this last constraint, sub-linear regret can be expected only if the graph presents some underlying structure. This paper shows that sub-linear regret is achievable in the case where the graph is generated according to a Stochastic Block Model (SBM) with two communities. Optimal regret bounds are computed for this pair-matching problem. They exhibit a phase transition related to the Kesten-Stigum threshold for community detection in SBM. The pair-matching problem is considered in the case where each node is constrained to be sampled less than a given amount of times. We show how optimal regret rates depend on this constraint. The paper is concluded by a conjecture regarding the optimal regret when the number of communities is larger than 2. Contrary to the two communities case, we argue that a statistical-computational gap would appear in this problem.
Fichier principal
Vignette du fichier
1905.07342v3.pdf (788.11 Ko) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-04578273 , version 1 (16-05-2024)

Identifiers

Cite

Christophe Giraud, Yann Issartel, Luc Lehericy, Matthieu Lerasle. Pair-Matching: Link Prediction with Adaptive Queries. Mathematical Statistics and Learning, 2024. ⟨hal-04578273⟩
77 View
9 Download

Altmetric

Share

Gmail Mastodon Facebook X LinkedIn More