Catálogo de publicaciones - libros

In this paper, we present a modified fuzzy min-max neural network model and its application to feature analysis. In the model a hyperbox can be expanded without considering the hyperbox contraction process as well as the overlapping test. During the learning process, the feature distribution information is utilized to compensate the hyperbox distortion which may be caused by eliminating the overlapping area of hyperboxes in the contraction process. The weight updating scheme and the hyperbox expansion algorithm for the learning process are described. A feature analysis technique for pattern classification using the model is also presented. We define four kinds of relevance factors between features and pattern classes to analyze the saliency of the features in the learning data set.

A novel hybrid cluster-based self-organizing neuro-fuzzy system (HC-SONFS) is proposed for dynamic function approximation and prediction. With the mechanism of self-organization, fuzzy rules are generated in the form of clusters using the proposed self-organization method to achieve compact and sufficient system structure if the current structure of knowledge base is insufficient to satisfy the required performance. A hybrid learning algorithm combining the well-known random optimization (RO) and the least square estimation (LSE) is use for fast learning. An example of chaos time series for system identification and prediction is illustrated. Compared to other approaches, excellent performance of the proposed HC-SONFS is observed.

In this paper, we propose a new machine learning approach based on AFS (Axiomatic Fuzzy Sets) fuzzy logic, in attempt to provide a better model with interpretability. First, we will concisely present the AFS theory. Second, we will propose new membership functions for fuzzy sets and their logic operations. Third, we will design a new machine learning algorithm based on the new membership functions and their logic operations. This algorithm has two advantages. One is that it can mimic the human reasoning comprehensively and offers a far more flexible and effective means for the study of large-scale intelligent systems. Another is its simplicity in implementation and mathematical beauty in fuzzy theory. Finally, a credit data example is used to illustrate its effectiveness.

Evaluation of the spinal forces from kinematics data is very complicated because it involves the handling of relationship between kinematic variables and electromyography (EMG) responses, as well as the relationship between EMG responses and the forces. A recurrent fuzzy neural network (RFNN) model is proposed to establish the kinematics-EMG-force relationship and model the dynamics of muscular activities. The EMG signals are used as an intermediate output and are fed back to the input layer. Since the EMG signal is a direct reflection of muscular activities, the feedback of this model has a physical meaning. It expresses the dynamics of muscular activities in a straightforward way and takes advantage from the recurrent property. The trained model can then have the forces predicted directly from kinematic variables while bypassing the procedure of measuring EMG signals and avoiding the use of biomechanics model. A learning algorithm is derived for the RFNN.

In this work, a new heuristic algorithm, named Swarm Double-Tabu Search (SDTS), has been proposed. SDTS attempts to solve the problems of NP-hard combinatorial optimization effectively and efficiently. The particle swarm and the double-tabu strategies adopted in the SDTS algorithm have got excellent search result. Simulations on Traveling Salesman Problem (TSP) were performed, and the results compared to those obtained by neural network approaches were optimal or near optimal.

In this work, authors have developed a system which is capable of composing songs using Genetic Algorithms (GA) to evolve melody and rhythm. Each GA uses two Multilayer Perceptron (MLP) type artificial neural networks (ANN) to judge for the fitness of individuals in the population. MLPs are forward and backward sliding-window predictors trained on melody and rhythm extracted from songs of different genres. Separately evolved rhythms and melodies are dynamically mixed to obtain verses, which are then mixed into whole songs.

A novel classification method based on regression is proposed in this paper and then the equivalences of the classification and regression are demonstrated by using numerical experiments under the framework of support vector machine. The proposed algorithm implements the classification tasks by the way used in regression problems. It is more efficiently for multi-classification problems since it can classify all samples at a time. Numerical experiments show that the two classical machine learning problems (classification and regression) can be solved by the method conventionally used for the opposite problem and the proposed regression-based classification algorithm can classify all samples belonging to different categories concurrently with an agreeable precision.

The unified fuzzy 9-intersection model of topological relations can describe the uncertainty of topological relations introduced by the uncertainty of spatial data. In this article, first, the raster algorithm for computing fuzzy 9-intersection model is presented, and the vector algorithms for computing fuzzy 9-intersection model of point/point, point/line, point/region, line/line, line/region, region/region topological relations are also provided. Second, based on the software developed by us, the examples for computing fuzzy 9-intersection matrix between two crisp objects, between two fuzzy objects and between a crisp object and a fuzzy object are listed. The results and analysis show that the unified fuzzy 9-intersection model is effective to describe the uncertainty of topological relations.

The location selection of distribution is one of the most important decision issues for logistic managers. In order to encompass vagueness in decision data, a new fuzzy multiple criteria decision-making method is proposed to solve the distribution center selection problem under fuzzy environment. In the proposed method, the ratings of alternatives and the weights of the criteria are given in terms of linguistic variables which is in turns represented by triangular fuzzy numbers.

In this paper, six different approaches using genetic algorithms (GA) and/or simulated annealing (SA) with improved bottom left (I-BL) algorithm [1] were applied for solution of two dimensional non-guillotine cutting problems. As examples, test problems including 29 individual rectangular pieces were used [2]. Performances of hybrid approaches on solutions of cutting problems were compared. Due to combined global search feature of GA and local search feature of SA, the hybrid approach using GA and SA yields the best results for these problems.