Fuzzy neural networks (FNNs) and rough neural networks (RNNs) both have been hot research topics in the artificial intelligence in recent years The former imitates the human brain in dealing with problems, the other takes advantage of rough set theory to process questions uncertainly The aim of FNNs and RNNs is to process the massive volume of uncertain information, which is widespread applied in our life This article summarizes the recent research development of FNNs and RNNs (together called granular neural networks) First the fuzzy neuron and rough neuron is introduced; next FNNs are analysed in two categories: normal FNNs and fuzzy logic neural networks; then the RNNs are analysed in the following four aspects: neural networks based on using rough sets in preprocessing information, neural networks based on rough logic, neural networks based on rough neuron and neural networks based on rough-granular; then we give a flow chart of the RNNs processing questions and an application of classical neural networks based on rough sets; next this is compared with FNNs and RNNs and the way to integrate is described; finally some advice is given on development of FNNs and RNNs in future

Granular neural networks

The rough neural networks (RNNs) are the neural networks based on rough set and one kind of hot research in the artificial intelligence in recent years, which synthesize the advantage of rough set to process uncertainly question: attributes reduce by none information losing then extract rule, and the neural networks have the strongly fault tolerance, self-organization, massively parallel processing and self-adapted. So that RNNs can process the massively and uncertainly information, which is widespread applied in our life. This article summarizes the recent research development of RNNs. First introduce the theory of rough set and the rough neuron; next analyze the RNNs in following four aspects: the neural network based on using rough set in preprocessing information, the neural networks base on rough logic, the neural networks based on rough neuron and the neural networks based on rough-granular; then give the flow chart of the RNNs processing question and the application of the classical neural network based on rough set; last give some advice to the development of RNNs in the future.

Rough Neural Networks: A Review

Granular neural networks(GNNs) as a new calculation system structure based on Granular Computing(GrC) and artificial neural network can be able to deal with all kinds of granular information of the real world. This article has made the summary on the development and the present situation of GNNs. Firstly, it introduces the basic model of GrC: word calculation model based on fuzzy sets theory, rough sets model, granular computing model based on quotient space theory and so on, summarizes the research progress of fuzzy neural networks(FNNs) and rough neural networks(RNNs), then analyses the ensemble-based methods of GNNs, researches their meeting point of three main GrC methods, and finally points out the research and development direction of GNNs.

/pdf/research-and-development-of-granular-neural-networks-v0sobkkd59.pdf

Research and Development of Granular Neural Networks

Breast cancer is the second leading cause of cancer deaths in women, and it is the most common type of cancer prevalent among women. Detecting tumor using mammogram is a difficult task because of complexity in the image. This brings the necessity of creating automatic tools to find whether a tumor is present or not. In this paper, rough set theory (RST) is integrated with back-propagation network (BPN) to classify digital mammogram images. Basically, RST is used to handle more uncertain data. Mammogram images are acquired from MIAS database. Artifacts and labels are removed using vertical and horizontal sweeping method. RST has also been used to remove pectoral muscles and segmentation. Features are extracted from the segmented mammogram image using GLCM, GLDM, SRDM, NGLCM, and GLRM. Then, the features are normalized, discretized, and then reduced using RST. After that, the classification is performed using RNN. The experimental results show that the RNN performs better than BPN in terms of classification accuracy.

Mammogram Image Classification Using Rough Neural Network

An approach to designing reduced-order H_∞ controller for a class of linear continuous dynamic systems is presented on Genetic Algorithm basis. Necessary and sufficient conditions for existence of the controller are given by means of linear matrix inequality. Taking the rank restriction condition as an object function of genetic algorithm, the minimum order n_k of the controller and the relevant paired parameter (X,Y) of positive definite matrix are obtained by searching the object function globally. Then, the reduced-order H_∞ controllers are designed. In the genetic algorithm as above the code is float, the selection operator is rank-based fitness assignment and elitist model, the crossover operator is real valued recombination and the mutation operator is real mutation. The simulation results show that the proposed controller has the same control effect as all-order H_∞ controllers based on MATLAB.

Reduced-Order H_∞ Controllers Based on Genetic Algorithm

The rough set theory is a new mathematical tool to study vague and uncertain information, and is widely used in intelligent systems. In this paper, the basic ideas of rough set theory are introduced, and the notion of up and low approximation sets, attribute reduction, core and some extensions of rough set theory are also presented. Then the application of rough set theory in intelligent systems. The combination of rough sets with fuzzy sets, neural network and genetic algorithms is mainly reviewed, which gives new ideas and methods to solve the hard problems in intelligent control.

Rough set theory and its application in the intelligent systems

The problem of mixed reduced-order H2/Hinfin controllers design is discussed that a general approach is proposed to design mixed reduced-order H2/Hinfin controllers by genetic algorithms. The mixed reduced-order H2/Hinfin controllers are designed which have robustness and better system performance through Hinfin controllers were H2 optimization by GA. GA is used twice in reduced-order Hinfin control and H2 optimization. The code is float, the selection operator is rank-based fitness assignment and elitist model, the crossover operator is real valued recombination, the mutation operator is real mutation, the fitness have penalty constrain. The simulation results show that mixed reduced-order H2/Hinfin controllers have not only better Hinfin performance but also better H2 performance.

Wei Pan

Papers

Rough set theory and its application in the intelligent systems

Mixed reduced-order H 2 /H ∞ controllers based on improved adaptive genetic algorithm