José Antonio Gómez-Ruiz

A principal components analysis self-organizing map

We propose a new self-organizing neural model that performs principal components analysis. It is also related to the adaptive subspace self-organizing map (ASSOM) network, but its training equations are simpler. Experimental results are reported, which show that the new model has better performance than the ASSOM network.

Analysis of OWA operators in decision making for modelling the majority concept

The majority concept plays a main role in decision making processes where one of the main problems is to define a decision strategy which takes into account the individual opinions of the decision makers to produce an overall opinion which synthesizes the opinions of the majority of the decision makers. The reduction of the individual values into a representative value of majority is usually performed trough an aggregation process. The most common operator used in these processes is the OWA operator, in which the majority concept can be modelled using fuzzy logic and linguistic quantifiers. In this work the fusion processes and the semantic used for modelling the majority concept in the OWA operators are analyzed and compared in order to present different approach to obtain a feasible majority aggregation value for the decision making problem.

Estimation of missing judgments in AHP pairwise matrices using a neural network-based model

Selecting relevant features to make a decision and expressing the relationships between these features is not a simple task. The decision maker must precisely define the alternatives and criteria which are more important for the decision making process. The Analytic Hierarchy Process (AHP) uses hierarchical structures to facilitate this process. The comparison is realized using pairwise matrices, which are filled in according to the decision maker judgments. Subsequently, matrix consistency is tested and priorities are obtained by calculating the matrix principal eigenvector. Given an incomplete pairwise matrix, two procedures must be performed: first, it must be completed with suitable values for the missing entries and, second, the matrix must be improved until a satisfactory level of consistency is reached. Several methods are used to fill in missing entries for incomplete pairwise matrices with correct comparison values. Additionally, once pairwise matrices are complete and if comparison judgments between pairs are not consistent, some methods must be used to improve the matrix consistency and, therefore, to obtain coherent results. In this paper a model based on the Multi-Layer Perceptron (MLP) neural network is presented. Given an AHP pairwise matrix, this model is capable of completing missing values and improving the matrix consistency at the same time.

Invariant pattern identification by self-organising networks

Abstract A new method for shape identification is proposed. It is not affected by similarity transformations (scalings, translations and rotations). The procedure is based on a robust index that gives the deformation of a self-organising feature map (SOFM) when it is trained with the object to identify.

Reconstruction methods for AHP pairwise matrices

Habitually, decision-makers are exposed to situations that require a lot of knowledge and expertise. Therefore, they need tools to help them choose the best possible alternatives. Analytic hierarchical process (AHP) is one of those tools and it is widely used in many fields. While the use of AHP is very simple, there is a situation that becomes complex: the consistency of the pairwise matrices. In order to obtain the consistent pairwise matrix from the inconsistent one, reconstruction methods can be used, but they cannot guarantee getting the right matrix according to the judgments of the decision maker. This situation does not allow proper evaluation of methods reliability, i.e. it is not possible to obtain a reliable ranking of alternatives based on an inconsistent matrix. In this work, a new way to evaluate the reliability of matrix reconstruction methods is proposed. This technique uses a novel measure for alternatives ranking comparison (based on element positions and distances), which is introduced in order to compare several matrix reconstruction methods. Finally, in order to demonstrate the extensibility of this new reliability measure, two reconstruction methods based on bio-inspired models (a Genetic Algorithm and the Firefly Algorithm) are presented and evaluated by using the aforementioned reliability measure.

Self-Organizing Dynamic Graphs

We propose a new self-organizing neural model that considers a dynamic topology among neurons. This leads to greater plasticity with respect to the self-organizing neural network (SOFM). Theorems are presented and proved that ensure the stability of the network and its ability to represent the input distribution. Finally, simulation results are shown to demonstrate the performance of the model, with an application to colour image compression.

Expansive and Competitive Learning for Vector Quantization

In this paper, we develop a necessary and sufficient condition for a local minimum to be a global minimum to the vector quantization problem and present a competitive learning algorithm based on this condition which has two learning terms; the first term regulates the force of attraction between the synaptic weight vectors and the input patterns in order to reach a local minimum while the second term regulates the repulsion between the synaptic weight vectors and the inputs gravity center to favor convergence to the global minimum This algorithm leads to optimal or near optimal solutions and it allows the network to escape from local minima during training. Experimental results in image compression demonstrate that it outperforms the simple competitive learning algorithm, giving better codebooks.

Voxel-Based Neighborhood for Spatial Shape Pattern Classification of Lidar Point Clouds with Supervised Learning

Improving the effectiveness of spatial shape features classification from 3D lidar data is very relevant because it is largely used as a fundamental step towards higher level scene understanding challenges of autonomous vehicles and terrestrial robots. In this sense, computing neighborhood for points in dense scans becomes a costly process for both training and classification. This paper proposes a new general framework for implementing and comparing different supervised learning classifiers with a simple voxel-based neighborhood computation where points in each non-overlapping voxel in a regular grid are assigned to the same class by considering features within a support region defined by the voxel itself. The contribution provides offline training and online classification procedures as well as five alternative feature vector definitions based on principal component analysis for scatter, tubular and planar shapes. Moreover, the feasibility of this approach is evaluated by implementing a neural network (NN) method previously proposed by the authors as well as three other supervised learning classifiers found in scene processing methods: support vector machines (SVM), Gaussian processes (GP), and Gaussian mixture models (GMM). A comparative performance analysis is presented using real point clouds from both natural and urban environments and two different 3D rangefinders (a tilting Hokuyo UTM-30LX and a Riegl). Classification performance metrics and processing time measurements confirm the benefits of the NN classifier and the feasibility of voxel-based neighborhood.

Design and optimization of symmetric laminated composites using a variable neighbourhood search-based model

The development of society is still marked by the need for lighter and stronger structures. The materials that respond best to these needs are composite materials. Designing composite materials is difficult as it involves designing the geometry and their composition. Traditionally, the design tasks have been based on approximate methods; the possibility for creating composite materials is almost unlimited, characterization by testing is very expensive and it is difficult to apply the results to other contexts. This article proposes a variable neighbourhood search-based model for the design of symmetric laminated composites, a general encoding for the design of composites, an evaluation function that has taken into consideration cost and safety criteria in design, the neighbourhood structures and a set of local search operators. The proposed model has been applied to different real-world problems and the results have been compared with other well-known design methods.

Multi-layer Perceptrons for Voxel-Based Classification of Point Clouds from Natural Environments

This paper addresses classification of 3D point cloud data from natural environments based on voxels. The proposed model uses multi-layer perceptrons to classify voxels based on a statistic geometric analysis of the spatial distribution of inner points. Geometric features such as tubular structures or flat surfaces are identified regardless of their orientation, which is useful for unstructured or natural environments. Furthermore, the combination of voxels and neural networks pursues faster computation than alternative strategies. The model has been successfully tested with 3D laser scans from natural environments.