Nakaji Honda

A study on the modeling ability of the IDS method: A soft computing technique using pattern-based information processing

The ink drop spread (IDS) method is a modeling technique developed by algorithmically mimicking the information-handling processes of the human brain. This method has been proposed as a new approach to soft computing. IDS modeling is characterized by processing that uses intuitive pattern information instead of complex formulas, and it is capable of stable and fast convergences. This paper investigates the modeling ability of the IDS method based on three typical benchmarks. Experimental results demonstrated that the IDS method can handle various modeling targets, ranging from logic operations to complex nonlinear systems, and that its modeling performance is satisfactory in comparison with that of feedforward neural networks.

Classification performance of the IDS method based on the two-spiral benchmark

The ink drop spread (IDS) method is a modeling technique used in the active learning method (ALM), which is a new approach to soft computing. It is characterized by a modeling process which is based on computing that uses intuitive pattern information instead of complex formulas. It has been proved that the IDS method is capable of stable fast modeling for complex nonlinear targets. In this paper, the classification performance of the IDS method is investigated. The two-spiral problem is a popular classification benchmark, and it is difficult to achieve the perfect classification due to high nonlinearity. With regard to this benchmark the IDS method exhibited good performance in terms of the classification rate and learning speed. This paper also present two learning modes, one of which is effective in solving the two-spiral problem rapidly.

A high performance IDS processing unit for a new fuzzy-based modeling

This paper presents a hardware unit for modeling systems using the active learning method (ALM). The ALM, a new methodology of soft computing, has processing units called IDSs, which are tasked with extracting useful information from a system subject to modeling. In realizing the ALM in hardware, it is desirable in terms of processing nature, performance, and scalability to utilize dedicated hardware for IDS. A developed high performance IDS processing unit enables ALM-based modeling systems to increase real-time capabilities. The hardware implementation of IDS and performance test results of the hardware are reported, and a consideration of the redundancy of IDS processing units is described.

Performance of the IDS Method as a Soft Computing Tool

Performance factors such as robustness, speed, and tractability are important for the realization of practical computing systems. The aim of soft computing is to achieve these factors in practice by tolerating imprecision and uncertainty instead of depending on exact mathematical computations. The ink drop spread (IDS) method is a modeling technique that has been proposed as a new approach to soft computing. This method is characterized by a modeling process that uses image information without including complex formulas. In this study, the performance of the IDS method is investigated in terms of robustness, speed, and tractability, which are typical criteria that determine the importance of soft computing tools. Robustness is evaluated on the basis of noise tolerance and fault tolerance. Tractability is discussed from the viewpoints of interpretability and transparency. Based on comparative evaluations with artificial neural networks and fuzzy inference systems, this study demonstrates that the IDS method has superior capability to function as a soft computing tool.

Acquisition of control knowledge of nonholonomic system by Active Learning Method

In this paper, we propose the Active Learning Method, the method to acquire the control knowledge actively by the method of trial and error. In this method, the input-output information is collected for the control object by the method of trial and error, and the controller is constructed based on the information. In the Active Learning Method, the output is decided actively and the action result is evaluated, and the data with high evaluation are modeled. This modeled pattern information becomes the behavior policy optimized based on the evaluation. For this modeling, the method called Ink Drop Spread method (IDS) is used. In this system, the object system is modeled functionally from the data by the fuzzy-like processing. By using the model of bar gymnast, the learning simulation is done for the behavior policy, and we examine the validity of this method.

Analysis of multivariate medical data by face method

Abstract This paper describes the application of face pattern as a medium to illustrate to man complex computer-processed medical diagnosis. The case of a nephrotic syndrome was selected to show the mechanics of the study. First, the design of the face pattern was constructed and psychometrical experiments analyzing the resulting facial expressions were conducted. Next, the results obtained from the analysis of facial expressions based and constructed from the original findings of conventional medical methods were used to loosely predict the patients future conditions, and the same results were compared with those of statistical procedures to test the efficiency of the method. Finally, a separate application of the method was also performed to indicate a possible effect of corticosteroid on idiopathic nephrotic syndrome.

A Fast Structural Optimization Technique for IDS Modeling

The ink drop spread (IDS) method is a modeling technique that is proposed as a new paradigm of soft computing. In this method, the structure of models is determined by the partitioning of the input domain. In order to obtain a high-accuracy model, it is necessary to determine the optimal number of partitions, i.e., structural optimization must be performed. This paper proposes a structural optimization technique for IDS modeling. The IDS model comprises multiple processing units, each of which is a modeling engine that develops a feature of the target system in the form of an easily comprehensible image on a two-dimensional plane. The proposed technique performs structural optimization with a small number of searches by analyzing the image information generated in the processing units instead of evaluating the model error using validation data.

Fault Tolerance Comparison of IDS Models with Multilayer Perceptron and Radial Basis Function Networks

The ink drop spread (IDS) method is a modeling technique developed by algorithmically mimicking the information-handling processes of the human brain. This method is a new paradigm of soft computing. The structure of IDS models is similar to that of artificial neural networks: they comprise distributed processing units. The beneficial property of fault tolerance is obtained when such parallel processing networks are implemented with dedicated hardware. This paper compares the IDS models with multilayer perceptron and radial basis function networks in terms of fault tolerance. The experimental results based on stuck-at fault tests reveal that the IDS models possess good fault tolerance.

Hardware for a new fuzzy-based modeling system and its redundancy

The active learning method (ALM) is a new methodology of soft computing. The ALM has processing engines called IDSs, which are tasked with extracting useful information from a system subject to modeling. In realizing the hardware for ALM, it is desirable in terms of processing nature, performance, and robustness to utilize dedicated hardware for the IDS. A high performance IDS processing unit that meets these requirements was developed. In this paper, outlines of the new IDS hardware and performance test results are presented. This paper also deals with the fault tolerance of ALM systems.

A hardware design for a new learning system based on fuzzy concepts

This paper presents a hardware system that implements the active learning method (ALM), a methodology of soft computing. ALM has processing engines called IDS, which are tasked with extracting useful information from a system subject to modeling. In realizing ALM in hardware, it will be desirable in terms of processing nature, performance, and scalability to utilize dedicated hardware for IDS. This paper primarily describes the actual hardware design of an IDS module, and shows the findings of tests of an ALM hardware system that implemented this module.