Iterative Methods for Signal Reconstruction on Graphs

Calogero Vetro, Elena Toscano, Emanuele Brugnoli

Risultato della ricerca: Other

Abstract

In applications such as social, energy, transportation, sensor, and neuronal networks, big data naturally reside on the vertices of graphs. Each vertex stores a sample, and the collection of these samples is referred to as a graph signal. The product of the network graph with the time series graph is considered as underlying structure for the evolution through time of graph signal “snapshots”. The framework of signal processing on graphs [4] extends concepts and methodologies from classical discrete signal processing. The task of sampling and recovery is one of the most critical topics in the signal processing community.In this talk, we present some localized iterative methods, obtained by modifying the Marvasti algorithm [2] in classical signal processing, for interpolating graph signals from only a partial set of samples, both in vertex and time domain. Our methods are also compared with other recent algorithms [3, 5] in order to study rate of convergence and computational efficiency [1]. The experimental results demonstrate the effectiveness of the proposed reconstruction methods in real world datasets and noisy scenarios.
Lingua originaleEnglish
Numero di pagine2
Stato di pubblicazionePublished - 2016

Fingerprint

Signal reconstruction
Iterative methods
Signal processing
Computational efficiency
Time series
Sampling
Recovery
Sensors

Cita questo

Iterative Methods for Signal Reconstruction on Graphs. / Vetro, Calogero; Toscano, Elena; Brugnoli, Emanuele.

2016.

Risultato della ricerca: Other

@conference{e42f58a9f33b4a57b735de1bb458650d,
title = "Iterative Methods for Signal Reconstruction on Graphs",
abstract = "In applications such as social, energy, transportation, sensor, and neuronal networks, big data naturally reside on the vertices of graphs. Each vertex stores a sample, and the collection of these samples is referred to as a graph signal. The product of the network graph with the time series graph is considered as underlying structure for the evolution through time of graph signal “snapshots”. The framework of signal processing on graphs [4] extends concepts and methodologies from classical discrete signal processing. The task of sampling and recovery is one of the most critical topics in the signal processing community.In this talk, we present some localized iterative methods, obtained by modifying the Marvasti algorithm [2] in classical signal processing, for interpolating graph signals from only a partial set of samples, both in vertex and time domain. Our methods are also compared with other recent algorithms [3, 5] in order to study rate of convergence and computational efficiency [1]. The experimental results demonstrate the effectiveness of the proposed reconstruction methods in real world datasets and noisy scenarios.",
keywords = "--",
author = "Calogero Vetro and Elena Toscano and Emanuele Brugnoli",
year = "2016",
language = "English",

}

TY - CONF

T1 - Iterative Methods for Signal Reconstruction on Graphs

AU - Vetro, Calogero

AU - Toscano, Elena

AU - Brugnoli, Emanuele

PY - 2016

Y1 - 2016

N2 - In applications such as social, energy, transportation, sensor, and neuronal networks, big data naturally reside on the vertices of graphs. Each vertex stores a sample, and the collection of these samples is referred to as a graph signal. The product of the network graph with the time series graph is considered as underlying structure for the evolution through time of graph signal “snapshots”. The framework of signal processing on graphs [4] extends concepts and methodologies from classical discrete signal processing. The task of sampling and recovery is one of the most critical topics in the signal processing community.In this talk, we present some localized iterative methods, obtained by modifying the Marvasti algorithm [2] in classical signal processing, for interpolating graph signals from only a partial set of samples, both in vertex and time domain. Our methods are also compared with other recent algorithms [3, 5] in order to study rate of convergence and computational efficiency [1]. The experimental results demonstrate the effectiveness of the proposed reconstruction methods in real world datasets and noisy scenarios.

AB - In applications such as social, energy, transportation, sensor, and neuronal networks, big data naturally reside on the vertices of graphs. Each vertex stores a sample, and the collection of these samples is referred to as a graph signal. The product of the network graph with the time series graph is considered as underlying structure for the evolution through time of graph signal “snapshots”. The framework of signal processing on graphs [4] extends concepts and methodologies from classical discrete signal processing. The task of sampling and recovery is one of the most critical topics in the signal processing community.In this talk, we present some localized iterative methods, obtained by modifying the Marvasti algorithm [2] in classical signal processing, for interpolating graph signals from only a partial set of samples, both in vertex and time domain. Our methods are also compared with other recent algorithms [3, 5] in order to study rate of convergence and computational efficiency [1]. The experimental results demonstrate the effectiveness of the proposed reconstruction methods in real world datasets and noisy scenarios.

KW - --

UR - http://hdl.handle.net/10447/237731

M3 - Other

ER -