This paper proposes the application of a genetic fuzzy rule-based classification system (GFRBCS) for tissue characterization of intravascular ultrasound (IVUS) images. The presented approach follows the IVUS Virtual Histology (IVUS-VH) plaque characterization technique, whereby the plaque region is classified into four primary tissue types, namely, calcium, necrotic core, fibrous and fibro-fatty. In order to increase the discrimination between the classes, a rich set of textural features is derived at different scales, including first-order statistics, gray-level co-occurrence matrices, run-lengths, wavelets, local binary patterns (LBP) and local indicators of spatial association (LISA) features. The employed fuzzy classifier effectively exploits the provided information, producing accurate and highly interpretable classification models. The extensive experimental analysis performed highlights the advantages of the proposed scheme against existing methods of the literature.

A novel fuzzy decision tree is proposed in this paper (the FDT-support vector machine (SVM) classifier), where the node discriminations are implemented via binary SVMs. The tree structure is determined via a class grouping algorithm, which forms the groups of classes to be separated at each internal node, based on the degree of fuzzy confusion between the classes. In addition, effective feature selection is incorporated within the tree building process, selecting suitable feature subsets required for the node discriminations individually. FDT-SVM exhibits a number of attractive merits such as enhanced classification accuracy, interpretable hierarchy, and low model complexity. Furthermore, it provides hierarchical image segmentation and has reasonably low computational and data storage demands. Our approach is tested on two different tasks: natural forest classification using a QuickBird multispectral image and urban classification using hyperspectral data. Exhaustive experimental investigation demonstrates that FDT-SVM is favorably compared with six existing methods, including traditional multiclass SVMs and SVM-based binary hierarchical trees. Comparative analysis is carried out in terms of testing rates, architecture complexity, and computational times required for the operative phase.

FastSLAM is one of the state-of-the-art approaches to the Simultaneous Localization and Mapping (SLAM) problem. In this paper, a new SLAM method is proposed, called L-SLAM, which is a low dimension version of the FastSLAM family algorithms. Dimensionality reduction of the particle filter is proposed, achieving better accuracy with less or the same number of particles. Dimensionality reduction of this problem renders the algorithm suitable for high dimensionality problems, like 3-D SLAM where the L-SLAM can produce better results in less time. Unlike the FastSLAM algorithms that uses Extended Kalman Filters (EKF), the L-SLAM algorithm updates the particles using Kalman filters. A methodology of linearizing a planar SLAM problem of a rear drive car-like robot is presented. Experimental results based on real case scenarios using the Car Park datasets and simulated environment are presented . The advantages of the proposed method in comparison with the FastSLAM 1.0 and 2.0 methods in the planar SLAM problem are discussed.

In this paper, a new SLAM method is proposed, called L-SLAM. It is a low dimension version of the FastSLAM family algorithms. The proposed method reduces the dimensionality of the particle filter that FastSLAM algorithms use, while achieving better accuracy with less or the same number of particles. Dimensionality reduction of this problem is the key feature for high dimensionality problems, like 3-D SLAM where the L-SLAM can produce better results in less time. In contrast to the FastSLAM algorithms that uses Extended Kalman Filters (EKF), the L-SLAM algorithm updates the particles using linear Kalman filters. A methodology of linearizing a planar SLAM problem of a front drive car-like robot is presented. Experimental results on a simulated environment demonstrates the advantages of the proposed method in comparison with the FastSLAM 1.0 and 2.0 methods in a planar SLAM problem.

This paper deals with the issue of gradual classification of a multivariate sequence where the number of candidate time-series generators is significantly high. It proposes a prediction scheme that consists of two components: a hierarchical structure which organizes the time-series models and a decision maker tool that assigns and evolves a respective hierarchy of probabilities; the latter expresses the current beliefs as to what the best model is for every hierarchical level. Experimentation in the domain of video-based human action recognition exhibits the capacity of the proposed approach to achieve efficient knowledge representation and real-time performance.

In this paper, we present a novel approach for 3D facial expression recognition which is inspired by the advances of ant colony and particle swarm optimization (ACO and PSO respectively) in the field of data mining. Anatomical correspondence between faces is first established using a generic 3D face model which is deformed elastically to match the facial surfaces. Surface points are then used as a basis for classification according to a set of classification rules, which are discovered by an ACO/PSO-based rule discovery algorithm. The performance of the proposed algorithm has been evaluated on the BU-3DFEDB facial expression database where a total recognition rate of 92.3% was achieved.

This work describes a two-mode clustering hierarchical model capable of dealing with high dimensional data spaces. The algorithm seeks a transformed subspace which can represent the initial data, simplify the problem and possibly lead to a better categorization level. We test the algorithm on two hard classification problems, the phoneme and the pedestrian recognition; both are typical classification problems from real-life applications. Finally, the model is compared with many other algorithms.

We introduce a dynamic programming algorithm to perform linear segmentation of concatenated texts by global minimization of a segmentation cost function which consists of: (a) within-segment word similarity (expressed in terms of the generalized density of the text dotplot) and (b) prior information regarding segment length. Our algorithm is evaluated on two Greek text collections and proves that it outperforms several other algorithms because it performs global optimization of a global cost function.

This paper addresses the problem of wind speed prediction at a particular location in the urban area of Thessaloniki, Greece, based on a data set containing wind parameter values at two other different locations. The problem is formulated and processed, and clusters in the data domain are identified by means of the Fuzzy Lattice Neurocomputing (FLN) methodology which can be used on any data set with the structure of a lattice. The results of the specific application are presented comparatively to past work, involving neural networks, on the same data set, demonstrating both the performance and the accuracy of the method.

straightforward method for the construction of Lyapunov functions represented by neural networks is presented in this paper. The resulting neural networks are Lyapunov functions on the basis of which asymptotic stability or instability of a nonlinear system's equilibrium point can be mathematically proven. One of the main advantages of this method is that it works for any nonlinear system even when the number of state variables is large. Several different Lyapunov functions can be constructed for each system, including Lyapunov functions of quadratic form. This enables us to select the most appropriate function for a given problem.

We present a methodology by which multiword communication could one day be realized in a population of evolving Artificial Life agents. The method which combines genetic algorithms and artificial neural networks has the prospect of a gradual increase of communication complexity (multi-word communication) through a series of evolutionary phases that take place in tandem. This is a layered treatment of gradual increasing complexity thus making the framework suitable for computational investigations. We present an example in which agents evolve a shared lexicon that pertains to two concepts (direction and quantity) and address the issue of the ordering of the lexical forms of those concepts.

This paper presents an evolutionary communication model that makes use of neural networks and evolutionary computation. It is embedded in a simple two-dimensional artificial world. The work builds on the successful approach of Cangelosi (1997) but is designed with considerations of suitability for extensions of more elaborate communication.

In this paper, we explore some aspects of the problem of online unsupervised learning of a switching time series, i.e., a time series which is generated by a combination of several alternately activated sources. This learning problem can be solved by a two-stage approach: 1) separating and assigning each incoming datum to a specific dataset (one dataset corresponding to each source) and 2) developing one model per dataset (i.e., one model per source). We introduce a general data allocation (DA) methodology, which combines the two steps into an iterative scheme: existing models compete for the incoming data; data assigned to each model are used to refine the model. We distinguish between two modes of DA: in parallel DA, every incoming datablock is allocated to the model with lowest prediction error; in serial DA, the incoming datablock is allocated to the first model with prediction error below a prespecified threshold. We present sufficient conditions for asymptotically correct allocation of the data. We also present numerical experiments to support our theoretical analysis.

We investigate the potential of a microgenetic algorithm (MGA) as a generalized hill-climbing operator. Combining a standard GA with the suggested MGA operator leads to a hybrid genetic scheme GA-MGA, with enhanced searching qualities. The main GA performs global search while the MGA explores a neighborhood of the current solution provided by the main GA, looking for better solutions. The MGA operator performs genetic local search. The major advantage of MGA is its ability to identify and follow narrow ridges of arbitrary direction leading to the global optimum. The proposed GA-MGA scheme is tested against 13 different schemes, including a simple GA and GAs with different hill-climbing operators. Experiments are conducted on a test set including eight constrained optimization problems with continuous variables. Extensive simulation results demonstrate the efficiency of the proposed GA-MGA scheme. For the same number of fitness evaluations, GA-MGA exhibited a significantly better performance in terms of solution accuracy, feasibility percentage of the attained solutions, and robustness.

A connectionist scheme, namely, /spl sigma/-Fuzzy Lattice Neurocomputing scheme or /spl sigma/-FLN for short, which has been introduced in the literature lately for clustering in a lattice data domain, is employed for computing clusters of directed graphs in a master-graph. New tools are presented and used, including a convenient inclusion measure function for clustering graphs. A directed graph is treated by /spl sigma/-FLN as a single datum in the mathematical lattice of subgraphs stemming from a master-graph. A series of experiments is detailed where the master-graph emanates from a thesaurus of spoken language synonyms. The words of the thesaurus are fed to /spl sigma/-FLN in order to compute clusters of semantically related words, namely hyperwords. The arithmetic parameters of /spl sigma/-FLN can be adjusted so as to calibrate the total number of hyperwords computed in a specific application. It is demonstrated how the employment of hyperwords implies a reduction, based on the a priori knowledge of semantics contained in the thesaurus, in the number of features to be used for document classification. In a series of comparative experiments for document classification, it appears that the proposed method favorably improves classification accuracy in problems involving longer documents, whereas performance deteriorates in problems involving short documents.

A basis for rigorous versatile learning is introduced theoretically, that is the framework of fuzzy lattices or FL-framework for short, which proposes a synergetic combination of fuzzy set theory and lattice theory. A fuzzy lattice emanates from a conventional mathematical lattice by fuzzifying the inclusion order relation. Learning in the FL-framework can be effected by handling families of intervals, where an interval is treated as a single entity/block the way explained here. Illustrations are provided in a lattice defined on the unit-hypercube where a lattice interval corresponds to a conventional hyperbox. A specific scheme for learning by clustering is presented, namely /spl sigma/-fuzzy lattice learning scheme or /spl sigma/-FLL (scheme) for short, inspired from adaptive resonance theory (ART). Learning by the /spl sigma/-FLL is driven by an inclusion measure /spl sigma/ of the corresponding Cartesian product to be introduced here. We delineate a comparison of the /spl sigma/-FLL scheme with various neural-fuzzy and other models. Applications are shown to one medical data set and two benchmark data sets, where /spl sigma/-FLL's capacity for treating efficiently real numbers as well as lattice-ordered symbols separately or jointly is demonstrated. Due to its efficiency and wide scope of applicability the /spl sigma/-FLL scheme emerges as a promising learning scheme.

Stapedotomy is a surgical procedure aimed at the treatment of hearing impairment due to otosclerosis. The treatment consists of drilling a hole through the stapes bone in the inner ear in order to insert a prosthesis. Safety precautions require knowledge of the nonmeasurable stapes thickness. The technical goal herein has been the design of high-level controls for an intelligent mechatronics drilling tool in order to enable the estimation of stapes thickness from measurable drilling data. The goal has been met by learning a map between drilling features, hence no model of the physical system has been necessary. Learning has been achieved as explained in this paper by a scheme, namely the d-sigma Fuzzy Lattice Neurocomputing (d sigma-FLN) scheme for classification, within the framework of fuzzy lattices. The successful application of the d sigma-FLN scheme is demonstrated in estimating the thickness of a stapes bone "on-line" using drilling data obtained experimentally in the laboratory.

The two-level fuzzy-lattice (2L-FL) learning scheme is introduced for application on an intelligent surgical (mechatronic) drill in the stapedotomy surgical procedure in the ear. The 2L-FL scheme learned from past cases to evaluate pointedly the thickness of a stapes bone using a force/torque pair of drilling profiles. Hence it is possible, in principle, to drill safely a hole through supple stapes by retracting automatically the drill upon bone breakthrough. The 2L-FL scheme was applied on two different partly ordered sets defined on a set of square matrices mapping features of two different data profiles to one another. Results are presented comparatively on experimental drilling data. The authors also discuss extension of their techniques to other surgical and industrial applications especially ones involving fusion of disparate sensory data where effective learning could compensate for the lack of accurate analytical models.

This paper proposes two hierarchical schemes for learning, one for clustering and the other for classification problems. Both schemes can be implemented on a fuzzy lattice neural network (FLNN) architecture, to be introduced herein. The corresponding two learning models draw on adaptive resonance theory (ART) and min-max neurocomputing principles but their application domain is a mathematical lattice. Therefore they can handle more general types of data in addition to N-dimensional vectors. The FLNN neural model stems from a cross-fertilization of lattice theory and fuzzy set theory. Hence a novel theoretical foundation is introduced in this paper, that is the framework of fuzzy lattices or FL-framework, based on the concepts fuzzy lattice and inclusion measure. Sufficient conditions for the existence of an inclusion measure in a mathematical lattice are shown. The performance of the two FLNN schemes, that is for clustering and for classification, compares quite well with other methods and it is demonstrated by examples on various data sets including several benchmark data sets.

We introduce a hybrid neural-genetic multimodel parameter estimation algorithm. The algorithm is applied to structured system identification of nonlinear dynamical systems. The main components of the algorithm are 1) a recurrent incremental credit assignment (ICRA) neural network, which computes a credit function for each member of a generation of models and 2) a genetic algorithm which uses the credit functions as selection probabilities for producing new generations of models. The neural network and genetic algorithm combination is applied to the task of finding the parameter values which minimize the total square output error: the credit function reflects the closeness of each model's output to the true system output and the genetic algorithm searches the parameter space by a divide-and-conquer technique. The algorithm is evaluated by numerical simulations of parameter estimation for a planar robotic manipulator and a waste water treatment plant.

We present a specific varying fitness function technique in genetic algorithm (GA) constrained optimization. This technique incorporates the problem's constraints into the fitness function in a dynamic way. It consists of forming a fitness function with varying penalty terms. The resulting varying fitness function facilitates the GA search. The performance of the technique is tested on two optimization problems: the cutting stock, and the unit commitment problems. Also, new domain-specific operators are introduced. Solutions obtained by means of the varying and the conventional (nonvarying) fitness function techniques are compared. The results show the superiority of the proposed technique.

A predictive modular neural network method is applied to the problem of unsupervised time-series segmentation. The method consists of the concurrent application of two algorithms: one for source identification, the other for time-series classification. The source identification algorithm discovers the sources generating the time series, assigns data to each source, and trains one predictor for each source. The classification algorithm recursively computes a credit function for each source, based on the competition of the respective predictors, according to their predictive accuracy; the credit function is used for classification of the time-series observation at each time step. The method is tested by numerical experiments.

We introduce the so-called predictive modular fuzzy system (PREMOFS) which performs time-series classification. A PREMOFS consists of 1) a bank of prediction modules and 2) a fuzzy decision module. It is assumed that the time series is generated by a source belonging to a finite search set (universal set); then the classification problem is to select the source that best represents the observed data, Classification is based on a membership function which is updated recursively according to the predictive accuracy of each model. Two algorithms are presented for updating the membership function. The first is based on sum/product fuzzy inference and the second on max/min fuzzy inference. In short, PREMOFS is a fuzzy modular system that classifies time series to one of a finite number of classes using the full set of past data (without preprocessing) to perform a recursive competitive computation of membership function based on predictive accuracy. Convergence proofs are given for both PREMOFS algorithms; in both cases the membership grade tends to one for the source that best predicts the observed data and to less than one for the remaining sources; hence, correct classification is guaranteed. Simulation results are also presented: PREMOFS are applied to signal detection, system identification, and phoneme classification tasks.

An incremental credit assignment (ICRA) scheme is introduced and applied to time series classification. It has been inspired from Bayes' rule, but the Bayesian connection is not necessary either for its development or proof of its convergence properties. The ICRA scheme is implemented by a recurrent, hierarchical, modular neural network, which consists of a bank of predictive modules at the lower level, and a decision module at the higher level. For each predictive module, a credit function is computed; the module that best predicts the observed time series behavior receives highest credit. We prove that the credit functions converge (with probability one) to correct values. Simulation results are also presented.

We apply the partition algorithm to the problem of time-series classification. We assume that the source that generates the time series belongs to a finite set of candidate sources. Classification is based on the computation of posterior probabilities. Prediction error is used to adaptively update the posterior probability of each source. The algorithm is implemented by a hierarchical, modular, recurrent network. The bottom (partition) level of the network consists of neural modules, each one trained to predict the output of one candidate source. The top (decision) level consists of a decision module, which computes posterior probabilities and classifies the time series to the source of maximum posterior probability. The classifier network is formed from the composition of the partition and decision levels. This method applies to deterministic as well as probabilistic time series. Source switching can also be accommodated. We give some examples of application to problems of signal detection, phoneme, and enzyme classification. In conclusion, the algorithm presented here gives a systematic method for the design of modular classification networks. The method can be extended by various choices of the partition and decision components.

This paper presents a robot system incorporating vision and force/torque sensing for handling of flat textile materials. Experimentation is used to draw conclusions concerning the performance of standard arms and sensing techniques. Also requirements of such systems are discussed.

This paper investigates the properties of the so-called feedforward method, which is a very simple training law suitable for on-chip learning. Its merit is conceptual and implementational simplicity. Its signals do not propagate in both directions and it works for various types of activation function, a feature that makes it particularly effective in the case of unmodeled activation functions. Extensive simulation has shown that this method is usually faster than backpropagation.