Multi-layer adaptive filters trained with back propagation: A statistical approach
| Autor: | Zakariya Faraj, Francis Castanie, Jean Claude Hoffmann, Mohamed Ibn Kahla |
|---|---|
| Rok vydání: | 1994 |
| Předmět: |
Signal processing
Steady state Artificial neural network Mean squared error business.industry Computer science Filter (signal processing) Perceptron Backpropagation Adaptive filter Least mean squares filter Control and Systems Engineering Multilayer perceptron Signal Processing Computer Vision and Pattern Recognition Artificial intelligence Electrical and Electronic Engineering business Algorithm Software Active noise control |
| Zdroj: | Signal Processing. 40:65-85 |
| ISSN: | 0165-1684 |
| DOI: | 10.1016/0165-1684(94)90022-1 |
| Popis: | This paper studies the convergence properties of a multi-layer linear neural network (MLLNN) within the framework of adaptive FIR filtering. The back propagation (BP) algorithm is used to adjust the network weights. We apply this structure to adaptive noise cancelling. The BP algorithm shows better performance than the classical adaptive line enhancer (ALE) trained with the LMS algorithm. The stability conditions, convergence speed, and steady state mean squared error are studied for a not fully connected network. The paper shows, in particular, the influence of the number of layers on the BP behaviour. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|