CS Talk - Rana Ali AMJAD : "Information Theoretic Cost Functions for Markov Aggregation and Clustering"

Rana Ali AMJAD - PhD at theTechnical University of Munich
Communication systems

Date: October 6th 2017
Location: Eurecom - Eurecom

Room Fourier (432), level - 4 Abstract : Markov aggregation is the task of representing a Markov chain with a large alphabet by a Markov chain with a smaller alphabet, thus reducing model complexity while at the same time retaining the computationally and analytically desirable Markov property. In this work we propose an information theoretic cost function for Markov aggregation. This cost function is motivated by two objectives: 1. The process obtained by observing the Markov chain through the mapping should be close to a Markov chain, and 2. the aggregated Markov chain should retain as much of the temporal dependence structure of the original Markov chain as possible. We then adapt the framework such that we can use it for clustering and co-clustering. This adapted framework includes the most famous co-clustering cost functions including ITCC and IBCC as special cases. Using this framework we then try to ascertain if information theoretic cost functions (proposed so far in the literature and the ones proposed by us as well) are a good choice for clustering problems. Bio : Rana Ali Amjad received his Bachelors degree in Electrical Engineering (with highest distinction) from University of Engineering and Technology, Lahore, Pakistan in 2011. He completed his Masters degree in Communication Engineering (with highest distinction) from Technical University of Munich, Germany in 2013. Since 2014 he is pursuing his PhD at the Institute for Communication Engineering at Technical University of Munich. He has received various awards in his academic career including the faculty award for best Master thesis, award for outstanding performance in Master’s degree and Gold medal for best performance in Communications major during his Bachelors degree. In his master thesis he focussed on developing the fundamental theory and coding techniques for Distribution Matching and Probabilistic Amplitude Shaping. At the beginning of his PhD he worked on different coding problems in Physical layer Security. Currently his main interest is on the interplay between Information/Coding theory and Machine Learning.

Permalink: https://www.eurecom/seminar/63371