Learnability of Competitive Threshold Models

TL;DR

This paper investigates the theoretical learnability of competitive threshold models for social contagion spread, demonstrating their simulation via neural networks and proposing efficient algorithms with practical applications.

Contribution

It introduces a novel theoretical framework for learning competitive threshold models, linking them to neural networks and providing sample complexity bounds.

Findings

Models can be simulated by neural networks with finite VC dimension.

Proposed algorithms perform well with limited data.

Method outperforms existing approaches on synthetic and real datasets.

Abstract

Modeling the spread of social contagions is central to various applications in social computing. In this paper, we study the learnability of the competitive threshold model from a theoretical perspective. We demonstrate how competitive threshold models can be seamlessly simulated by artificial neural networks with finite VC dimensions, which enables analytical sample complexity and generalization bounds. Based on the proposed hypothesis space, we design efficient algorithms under the empirical risk minimization scheme. The theoretical insights are finally translated into practical and explainable modeling methods, the effectiveness of which is verified through a sanity check over a few synthetic and real datasets. The experimental results promisingly show that our method enjoys a decent performance without using excessive data points, outperforming off-the-shelf methods.