Fuzzy perceptron learning and its application to classifiers with numerical data and linguistic knowledge
Jialin Chen, Jyh‐Yeong Chang · 2002
This paper proposes a fuzzy perceptron neural network learning algorithm for classifiers that use expert knowledge represented by fuzzy if-then rules as well as numerical data. We extend the conventional linear perceptron network to a second-order one that provides much more discrimination flexibility. In order to handle linguistic variables in neural networks, fuzzy set levels are incorporated into perceptron neural learning. At different levels of the input fuzzy number, the fuzzy perceptron algorithm is derived from the fuzzy output function and the corresponding nonfuzzy target output that indicates the correct class of the fuzzy input vector. Moreover, the pocket algorithm is modified according to our fuzzy perceptron learning scheme and called the fuzzy pocket algorithm, to solve nonseparability problems, such as overlapping fuzzy inputs. Simulation results are provided to demonstrate the power of the proposed algorithm.