Accession Number:

AD1020605

Title:

Multi-modal Social Networks: A MRF Learning Approach

Descriptive Note:

Technical Report,01 Sep 2011,31 Aug 2014

Corporate Author:

University of Texas at Austin Austin United States

Personal Author(s):

Report Date:

2016-06-20

Pagination or Media Count:

9.0

Abstract:

The work primarily focused on two lines of research.1. We propose new greedy algorithms for learning the structure of a graphical model of a probability distribution, given samples drawn fromthe distribution. Our research modifies greedy algorithms through appropriate node pruning, to result in fast algorithms that provideanalytical guarantees on correctness.2. The objective of this line of work is to use noisy measurements from cascades stochastic processes for spread on graphs to learn thespread of information opinion malware. Our approach for this learning problem has been to view this as hypothesis testing on graphs given noisy and partial information on both node states and the network graph, we formulate the problem as distinguishing between a benignhypothesis no spreading process and a malicious hypothesis spreading process such as malware. This approach has been used in asequence of studies, starting from distinguishing with partial information, to that with nodes with are adversarial nodes could lie about theirstate, to dealing with noisy network knowledge. We have also been able to use this approach to learn the identity of communities withshared interests.

Subject Categories:

  • Statistics and Probability
  • Operations Research
  • Sociology and Law
  • Computer Systems

Distribution Statement:

APPROVED FOR PUBLIC RELEASE