 |
|
The ISIS Group introduced the neurofuzzy modelling algorithm in 1989 that incorporated the learning properties of associative memory neural networks, yet had the interpretation of fuzzy logic. Subsequently the algorithm has been refined to overcome the curse of dimensionality through a variety of schemas including adaptive spline modelling (ASMOD), extended additive modelling (to accommodate multiplicative processes), the Delaunay triangle algorithms and more recently the Bezier-Bernstein Algorithm that incorporates all the properties of the B-spline algorithm yet is parsimonious. Other modelling approaches include a hybrid mixture of expert algorithm and a multi-resolution (akin to wavelets) representation which provide alternatives to RBF modelling of unknown processes. Applications include materials classification, sensors, car engine modelling, and conditioning monitoring in power generation. |
