Linguistic rule extraction from neural networks and genetic-algorithm-based rule selection

Abstract

Abstract-We present an approach for automated evolutionary design of the functionary of driving agent, able to operate a software model of fast running car. The objective of our work is to automatically discover a set of driving rules (if existent) that are general enough to be able to adequately control the car in all sections of predefined circuits. In order to evolve an agent with such capabilities, we propose an indirect, generative representation of the driving rules as algebraic functions of the features of the current surroundings of the car. These functions, when evaluated for the current surrounding of the car yield concrete values of the main attributes of the driving style (e.g., straight line velocity, turning velocity, etc.), applied by the agent in the currently negotiated section of the circuit. Experimental results verify both the very existence of the general driving rules and the ability of the employed genetic programming framework to automatically discover them. The evolved driving rules offer a favorable generality, in that a single rule can be successfully applied (i) not only for all the section of a particular circuit, but also (ii) for the sections in several a priori defined circuits featuring different characteristics.

Designing simulated annealing and subtractive clustering based fuzzy classifier

“…In order to guarantee in finding global optima, our modified algorithm searches the new configuration in all feasible regions with a transition function, shown in Eq. (16), in high temperature values in which acceptance probability is high with taking a worst solution in order to escape global minima. In low temperature values, algorithm searches a broader range according to high temperature with searching neighborhood of current states, as described in Eq.…”

“…In the next step, new configuration is generated using either Eq. (15) or (16). The classifier is trained and tested with generated configuration.…”

“…In this situation, some systematic ways have been proposed in order to construct fuzzy if-then rule base [10][11][12][13]. Some works have been made on how fuzzy system could be extracted from numerical data with genetic algorithm [14][15][16], tabu search [17,18], decision trees [19] and evolutionary programming [20,21] techniques.…”

An interval type-2 fuzzy system with hybrid intelligent learning

“…Automatic construction methods of fuzzy rules from numerical data of particular domains have been extensively investigated. Artificial neural networks (ANNs) [9], genetic algorithms (GAs) [10]- [13], Particle Swarm Optimization (PSO) [14], and other learning algorithms [6][15]- [17] have been used to aid the generation of fuzzy rules. Clustering algorithms such as kmeans, fuzzy c-means, and self-organizing feature map algorithms have been employed to assist fuzzy rule generation.…”

“…Ishibuchi et al [10] used the genetic algorithm to derive fuzzy rules from numerical data. Using this method, a compact rule set is obtained since several unnecessary fuzzy partitions are removed by the genetic operations.…”