Polynomial Neuro-Fuzzy Architectures:Focused On Modeling And Application

With the development of the intelligent technologies, the conventional artificial intelligence viewpoint that is based on the signal rule formatted from human language have been tranformed into the current computational intelligence (CI) that is based on the fusion of fuzzy models, neural networks, and evolutionary algorithms. In this thesis, we have proposed a serial polynomial neural fuzzy models based on CI. The overall work includes mainly in the following aspects:First, we have proposed polynomial fuzzy models with two polynomial order. Two different polynomial fuzzy models, namely SA fuzzy models and MA fuzzy models, have been developed based on different fuzzy rules.Second, we have proposed polynomial fuzzy neural networks. Four polynomial fuzzy neural networks whose basic fuzzy neurons of networks are essentially the polynomial fuzzy models have been designed. They are as follows:FPSAFNN、 HPSAFNN、FPMAFNN、HPMAFNN.Third, we have proposed SPOA optimized two order polynomial fuzzy models. An optimized algorithm, namely SPOA, have been proposed to optimize the polynomila fuzzy models. The SPOA optimized fuzzy models are as follows:SPOA optimized SA fuzzy models and SPOA optimized MA fuzzy models.Fourth, we have proposed SPOA optimized polynomial fuzzy neural networks. We have deigned four networks that are as follows:SPOA optimized FPSAFNN, SPOA optimized HPSAFNN, SPOA optimized FPMAFNN, SPOA optimized HPMAFNN.At last, we have employed the proposed polynomial neural fuzzy models to natural language processing. We take a classification problem in natural language processing as an example.