Abstract | ||
---|---|---|
There have been several interesting attempts to blend fuzzy set logic and neural networks. These include Shiue and Grondin [3], Yager [4], Taber and Deich [7], Kosko [10], Oden [8], and more recently Yamakawa and Tomoda [1] and Langheld and Goser [9]. This paper extends the work of Yamakawa and Tomoda by employing Maximum Entropy Ordered Weighted Averaging (MEOWA) operations at the soma of a fuzzy neuron. The MEOWA is an extension by O'Hagan [6, 2] of Yager's [4, 5] Ordered Weighted Averaging (OWA) operators. The advantage of using the MEOWA operator in a fuzzy neuron is that a single type IC appears to be sufficient to build a neural pattern recognition system. Neural inputs and weights are fuzzy numbers rather than the usual scalar values of "standard" neural networks. Fuzzy logic operations replace the usual summing, dot product, and nonlinear "squashing" operations in the soma. Fuzzy neurons appear to be able to perform pattern recognition with reduced resolution inputs relative to standard neural networks: O(m) + O(n) vs. O(mn). |
Year | DOI | Venue |
---|---|---|
1990 | 10.1007/BFb0028148 | IPMU |
Keywords | DocType | ISBN |
maximum entropy,weighted averaging,fuzzy neuron,neural network,fuzzy logic,pattern recognition,fuzzy number,fuzzy set | Conference | 3-540-54346-5 |
Citations | PageRank | References |
8 | 1.14 | 2 |
Authors | ||
1 |
Name | Order | Citations | PageRank |
---|---|---|---|
Michael O'Hagan | 1 | 8 | 1.14 |