Kazuyuki Murase

In this paper, an integration of fuzzy logic controllers (FLCs) with hybrid genetic algorithms (HGAs) is developed with a view to make the design process fully automatic, without requiring any human expert and numerical data. Our approach consists of two phases: first phase involves selection and definition of fuzzy control rules as well as adjustment of membership functions parameters, while the second phase performs an optimal selection of membership function types corresponding to fuzzy control rules. Learning both parts concurrently represents a way to improve the accuracy of the FLCs to minimize the errors. It has been argued that the performance of FLCs greatly depends on the parameters as well as types of membership functions. Thus, the aforementioned HGAs are a viable solution for designing an efficient adaptive FLCs system. To demonstrate the effectiveness of the proposed method for optimal design of the FLCs, the proposed approach is applied to a well-known benchmark controller design tasks, car and truck-and-trailer like robot system. Simulation results illustrates that proposed optimization approach can find optimal fuzzy rules and their corresponding membership functions types with a high rate of accuracy. The new HGAs optimized adaptive FLCs outperforms not only a passive control strategy but also human-designed FLCs, a neural coded controller with clustering and a neural-fuzzy control algorithm.

Capacitated Vehicle Routing Problem (CVRP) is anoptimization task where customers are assigned to vehicles aiming that combined travel distances of all the vehicles as minimum as possible while serving customers. A popular way among various methods of CVRP is solving it in two phases: grouping or clustering customers into feasible routes of individual vehicles and then finding their optimal routes. Sweep is well studied clustering algorithm for grouping customers and different traveling salesman problem (TSP) solving methods are commonly used to generate optimal routes of individual vehicles. This study investigates effective CVRP solving method based on recently developed adaptive Sweep and prominent Swarm Intelligence (SI) based TSP optimization methods. The adaptive Sweep cluster is a heuristic based adaptive method to select appropriate cluster formation starting angle of Sweep. Three prominent SI based TSP optimization methods are investigated which are Ant Colony Optimization, Producer-Scrounger Method and Velocity Tentative Particle Swarm Optimization (VTPSO). Genetic Algorithm is also considered since it is the pioneer and well-known population based method. The experimental results on two suites of benchmark CVRPs identified the effectiveness of adaptive Sweep plus SI methods in solving CVRP. Finally, adaptive Sweep plus the VTPSO is found better than other tested methods in this study as well as several other prominent existing methods.

Image classification, a complex perceptual task with many real life important applications, faces a major challenge in presence of noise. Noise degrades the performance of the classifiers and makes them less suitable in real life scenarios. To solve this issue, several researches have been conducted utilizing denoising autoencoder (DAE) to restore original images from noisy images and then Convolutional Neural Network (CNN) is used for classification. The existing models perform well only when the noise level present in the training set and test set are same or differs only a little. To fit a model in real life applications, it should be independent to level of noise. The aim of this study is to develop a robust image classification system which performs well at regular to massive noise levels. The proposed method first trains a DAE with low-level noise-injected images and a CNN with noiseless native images independently. Then it arranges these two trained models in three different combinational structures: CNN, DAE-CNN, and DAE-DAE-CNN to classify images corrupted with zero, regular and massive noises, accordingly. Final system outcome is chosen by applying the winner-takes-all combination on individual outcomes of the three structures. Although proposed system consists of three DAEs and three CNNs in different structure layers, the DAEs and CNNs are the copy of same DAE and CNN trained initially which makes it computationally efficient as well. In DAE-DAE-CNN, two identical DAEs are arranged in a cascaded structure to make the structure well suited for classifying massive noisy data while the DAE is trained with low noisy image data. The proposed method is tested with MNIST handwritten numeral dataset with different noise levels. Experimental results revealed the effectiveness of the proposed method showing better results than individual structures as well as the other related methods.

Convolutional neural network (CNN) has been successfully used in many fields including image recognition. CNN is composed of input, convolution, pooling, hidden and output layers, and the weights and biases between layers except the ones between convolution and pooling layers are acquired by learning. In comparison to the conventional neural networks, the learning cost of CNN is higher, and the learning time is longer especially when hidden layer(s) are added. Recently, complex- and quaternion-valued neural networks have drawn much attention. In complex-valued neural networks, inputs, weights, biases and outputs are complex numbers, and in quaternion-valued neural networks, these parameters are quaternions. It has been shown that both methods exhibit excellent accuracy in various applications such as classification and function approximation problems with less computational burden. In this study, we propose CNNs with complex- and quaternion-valued neurons where complex and quaternion numbers are used between pooling, hidden and output layers. We here show that CNNs with complex- and quaternion-valued neurons have higher learning ability in handwritten digit image classification with the MNIST data than the real-valued counterpart.

Vehicle Routing Problem (VRP) has become an integral part in logistic operations which determines optimal routes for several vehicles to serve customers. The basic version of VRP is Capacitated VRP (CVRP) which considers equal capacities for all vehicles. The objective of CVRP is to minimize the total traveling distance of all vehicles to serve all the customers. Various methods are used to solve CVRP, among them the most popular way is splitting the task into two different phases: assigning customers under different vehicles and then finding optimal route of each vehicle. Sweep clustering algorithm is well studied for clustering nodes. On the other hand, route optimization is simply a traveling salesman problem (TSP) and a number of TSP optimization methods are applied for this purpose. In Sweep, cluster formation staring angle is identified as an element of CVRP performance. In this study, a heuristic approach is developed to identify appropriate starting angle in Sweep clustering. The proposed heuristic approach considers angle difference of consecutive nodes and distance between the nodes as well as distances from the depot. On the other hand, velocity tentative particle swarm optimization (VTPSO), the most recent TSP method, is considered for route optimization. Finally, proposed adaptive Sweep (i.e., Sweep with proposed heuristic) plus VTPSO is tested on a large number of benchmark CVRP problems and is revealed as an effective CVRP solving method while outcomes compared with other prominent methods.

Two-layered neural networks are well known as autoencoders (AEs) in order to reduce the dimensionality of data. AEs are successfully employed as pre-trained layers of neural networks for classification tasks. Most of the existing studies conceived real-valued AEs in real-valued neural networks. This study investigated complex- and quaternion-valued AEs for complex- and quaternion-valued neural networks. Inputs, weights, biases, and outputs in complex-valued AE (CAE) are complex variables, whereas those in quaternion-valued AE (QAE) are quaternions. In both methods, a split-type activation function is used in the hidden and output units. To deal with the images using the proposed methods, pairs of pixels are allotted to complex-valued inputs in the CAE and quartets of pixels are allotted to quaternion-valued inputs in the QAE. Proposed autoencoders are tested and performance compared with conventional AE for several tasks which are encoding/decoding, handwritten numeral recognition and large-scale multi-class classification. Proposed CAE and QAE revealed as good recognition methods for the tasks and outperformed conventional AE with significance performance in case of large-scale multi-class images recognition.

The Presence of unique restriction sites (URSs) within a sequence is very important so that the sequence may be cut unambiguously in exactly one place with a restriction enzyme. Restriction site manipulation in virus genome may produce virus variants to serve as potential vaccines. Therefore, a number of applications are invented for the manipulation of viral genomes to produce attenuated viruses. Recently, automatic generation of URSs in the sequence has been investigated through different approaches (e.g., Greedy, Weighted Bipartite and Max-Min Gap) and found effective. The aim of this study is to investigate URS enhancement in a given virus genome considering it as an optimization task. URS placement (URSP) is a kind of pattern matching problem; therefore, formulation of existing optimization approach(s) to solve it efficiently is very important timely research so that working with large sized genomes becomes easy. In this study, Particle Swarm Optimization (PSO), the popular method of optimization, has been modified and formulated to solve URSP problem. The proposed URSP-PSO method has been tested on a set of benchmark virus genome sequences and found to increase URSs in a large number. The method is also found better than existing methods for large sized genome sequences.