March 16, 2025
Taher Yazdanpanah

Taher Yazdanpanah

Academic Rank: Associate professor
Address:
Degree: Ph.D in -
Phone: -
Faculty: Faculty of Intelligent Systems and Data Science

Research

Title
Refining Directed Graphs for Spectral Analysis through Strategic Topological Modifications
Type Presentation
Keywords
Graph Fourier transform, Directed graph, Eigendecomposition, Jordan form
Researchers Lidija Vuksanovic (First researcher) , Ali Bagheri-Bardi (Second researcher) , Milos Dakovic (Third researcher) , Taher Yazdanpanah (Fourth researcher) , Ljubiša Stanković (Fifth researcher)

Abstract

—In graph signal processing, the absence of a welldefined graph Fourier transform complicates spectral analysis on directed graphs. To address this, we propose a technique that subtly modifies the graph’s topology, making minimal changes to facilitate spectral analysis. Our approach starts by adding edges to eliminate sources and sinks, followed by identifying connections that impede the non-singularity and diagonalizability of the adjacency matrix. We then make minor edge weight adjustments to achieve diagonalizability, ensuring efficient spectral analysis of graph signals defined on directed graph.