A Modified Water Flow-like Algorithm for TSK-type Interval-valued Neural Fuzzy System Design and Its Application in Blind Source Separation
| Autor: | Che-Ting Kuo, 郭哲廷 |
|---|---|
| Rok vydání: | 2010 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 98 Based on the water flow-like algorithm (WFA), we propose a novel hybrid learning algorithm: the modified water flow-like algorithm (MWFA) for TSK-type interval-valued neural fuzzy system with asymmetric membership function (TIVNFS-A) design. The WFA is inspired from the natural behavior of the water flows. For finding the global optimum in the solution space, the splitting, moving, merging, evaporation, and precipitation operations have been adopted. The WFA has global search ability and has dynamic solution agents. However, the original WFA is not suited to the continuous solution space due to the strategies. Therefore, we propose the MWFA for the continuous solution space. We propose the novel moving strategies by applying the tabu search algorithm and BP to improve the performance and enhance the strategies in evaporation and precipitation operations. These enhanced strategies in evaporation and precipitation evaporation are more consistent with the natural behavior of water flows and increase the diversity of solution agents. Therefore, we use the MWFA to train the TIVNFS-A, identify the performance by the example of nonlinear system identification and design a novel approach for solving the blind source separation (BSS) problem. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|