Adaptive Basis Function Selection for Computationally Efficient Predictions
Uloženo v:
| Název: | Adaptive Basis Function Selection for Computationally Efficient Predictions |
|---|---|
| Autoři: | Kullberg, Anton, Viset, Frida, Skog, Isaac, Hendeby, Gustaf |
| Zdroj: | Wallenberg AI, Autonomous Systems and Software Program (WASP) IEEE Signal Processing Letters. 31:2130-2134 |
| Témata: | Computational modeling, Predictive models, Adaptation models, Data models, Accuracy, Training data, Probabilistic logic, Adaptive signal processing, computational complexity, function approximation, Gaussian processes, WASP_publications |
| Popis: | Basis Function (BF) expansions are a cornerstone of any engineer's toolbox for computational function approximation which shares connections with both neural networks and Gaussian processes. Even though BF expansions are an intuitive and straightforward model to use, they suffer from quadratic computational complexity in the number of BFs if the predictive variance is to be computed. We develop a method to automatically select the most important BFs for prediction in a sub-domain of the model domain. This significantly reduces the computational complexity of computing predictions while maintaining predictive accuracy. The proposed method is demonstrated using two numerical examples, where reductions up to 50-75% are possible without significantly reducing the predictive accuracy. |
| Popis souboru: | electronic |
| Přístupová URL adresa: | https://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-206535 https://liu.diva-portal.org/smash/get/diva2:1890083/FULLTEXT01.pdf |
| Databáze: | SwePub |
Full Text Finder 
Nájsť tento článok vo Web of Science | Abstrakt: | Basis Function (BF) expansions are a cornerstone of any engineer's toolbox for computational function approximation which shares connections with both neural networks and Gaussian processes. Even though BF expansions are an intuitive and straightforward model to use, they suffer from quadratic computational complexity in the number of BFs if the predictive variance is to be computed. We develop a method to automatically select the most important BFs for prediction in a sub-domain of the model domain. This significantly reduces the computational complexity of computing predictions while maintaining predictive accuracy. The proposed method is demonstrated using two numerical examples, where reductions up to 50-75% are possible without significantly reducing the predictive accuracy. |
|---|---|
| ISSN: | 10709908 15582361 |
| DOI: | 10.1109/lsp.2024.3445272 |