Evolutionary neural network learning algorithms for changing environments
Classical Machine Learning methods are usually developed to work in static data sets. Yet, real world data typically changes over time and there is the need to develop novel adaptive learning algorithms. In this work, a number of algorithms, combining Neural Network learning models and Evolutionary...
| Autor principal: | |
|---|---|
| Outros Autores: | , |
| Formato: | article |
| Idioma: | eng |
| Publicado em: |
2004
|
| Assuntos: | |
| Texto completo: | http://hdl.handle.net/1822/426 |
| País: | Portugal |
| Oai: | oai:repositorium.sdum.uminho.pt:1822/426 |
| Resumo: | Classical Machine Learning methods are usually developed to work in static data sets. Yet, real world data typically changes over time and there is the need to develop novel adaptive learning algorithms. In this work, a number of algorithms, combining Neural Network learning models and Evolutionary Computation optimization techniques, are compared, being held several simulations based on artificial and real world problems. The results favor the combination of evolution and lifetime learning according to the Baldwin effect framework. |
|---|