Adaptive Filtering Algorithms: New and Old
Autore
Giovanni Vecchiato - Università degli Studi di Napoli - Federico II - [2005-06]
Documenti
Abstract
Objective of this thesis work is analyzing and comparing two different types of adaptive systems. In particular, we focus our attention on behaviour of the Echo-State Network (ESN) and the Least Mean-Square algorithm (LMS). The former is a new born kind of Recurrent Neural Network while the latter represents the most used technique concerning adaptive filtering.
Our purpose is designing a filter for prediction using such as algorithms. After introducing motivations for using adaptive systems, we will explain theories related to ESN and LMS. Afterwards, we will show how to predict a signal with these two techniques and then their performance. Finally, we will illustrate basic differences between the algorithms and that LMS presents advantages both in performance and in computational time with respect to ESN.
Our purpose is designing a filter for prediction using such as algorithms. After introducing motivations for using adaptive systems, we will explain theories related to ESN and LMS. Afterwards, we will show how to predict a signal with these two techniques and then their performance. Finally, we will illustrate basic differences between the algorithms and that LMS presents advantages both in performance and in computational time with respect to ESN.
Questa tesi è correlata alla categoria