fuzzy inference system (fis)

Soft computing algorithm such as fuzzy logic, neural networks, and evolutionary algorithms are widely used in many fields. Fuzzy logic, in particular, has gained significant popularity due to its significant ability in modelling. So far, various methods of fuzzy modelling have been presented; each of these methods has its advantages and disadvantages. While all methods start from the input, Sugeno-Yasukawa (SY) differs by initiating the analysis from the output. The popularity of the SY method can be attributed to its effective rule extraction algorithm, which employs a clustering process to determine input membership functions. In this paper, we propose a cluster search algorithm and a new fuzzy partitioning method that enhance the mapping of the output space to the input space by distributing Gaussian functions for each data point within a cluster and calculating their membership values. With this proposed new clustering search method, the performance of the SY method is improved. Through simulations, the proposed method has improved the mean square errors (MSE) criterion by 0.001, and improved the accuracy criterion by 1.5 percent.

In the training phase of learning algorithms, it is always important to have a suitable training data set. The presence of outliers, noise data, and inappropriate data always affects the performance of existing algorithms. The active learning method (ALM) is one of the powerful tools in soft computing inspired by the computation of the human brain. The operation of this algorithm is completely based on simple calculations and tries to model a Multi-Input Single-Output (MISO) system as a set of Single-Input Single-Output (SISO) subsystems and breaks a complex problem into several simpler problems. Each of the subsystems is then modeled by an Ink Drop Spread (IDS) plane. In this paper, to improve the performance of ALM, with changes in how it works, data called negative data has been used. In the ALM, it is possible to use negative data in the training phase by using the IDS operator. By doing so, a policy similar to the reward and punishment policy is in reinforcement learning methods is used. To investigate the accuracy of the proposed algorithm, some simulations in control have been done on an inverted pendulum system and the FVU value is 0.0143. The simulations results confirm the proper performance and increase the computational power of the proposed method compared to the existing method.