General Notices

The Department of Mathematics and Statistics will host a Ph.D defense on Thursday, January 23 from 2:00 p.m. to 4:00 p.m. Ms. Trisha Lawrence will present on the topic   On The Stochastic Dynamic Programming Of Influence Maximization-Revenue Optimization

Interested persons are invited to attend virtually via Zoom. Click here to access (Meeting ID: 932 5296 2711 | Passcode: 617234).

Abstract:

This thesis centers around the influence that friends have on each other in Online Social Networks (OSNs) and explores a novel decision-making approach to the well-known Influence Maximization (IM) problem. This thesis diverts from previous approaches to the problem which originated from the term influence spread and adopt a novel approach geared towards optimizing revenue to the advertiser. Thus a new problem is defined as the Influence Maximization-Revenue Optimization (IM-RO) problem and Stochastic Dynamic Programming (SDP) is introduced as a method in which this problem can be solved. Due to the “curse of dimensionality” associated with implementing SDP, heuristic methods which obtain near-optimal solutions are proposed. Moreover, in this thesis, the intuition behind our proposed influence model for the IM-RO problem is validated and its model parameters estimated empirically. Further, an approximate iterative algorithm and Lagrangean relaxation with the subgradient method is adopted as an alternative to the heuristic methods for obtaining near-optimal solutions. The primary goal in this thesis is to present a SDP formulation to the well-known IM problem and propose methods for obtaining feasible, near-optimal solutions to large scale problems, involving OSNs.

Keywords — Stochastic dynamic programming; Influence maximization-revenue optimization; influence models; Lagrangean relaxation/ subgradient methods.