Search

The final publication is available at Springer via http://dx.doi.org/10.1007/11499305_19, Proceedings of First International Work-Conference on the Interplay Between Natural and Artificial Computation, IWINAC 2005, Las Palmas, Canary Islands, Spain, June 15-18, 2005, This paper describes Attribute Grammar Evolution (AGE), a new Automatic Evolutionary Programming algorithm that extends standard Grammar Evolution (GE) by replacing context-free grammars by attribute grammars. GE only takes into account syntactic restrictions to generate valid individuals. AGE adds semantics to ensure that both semantically and syntactically valid individuals are generated. Attribute grammars make it possible to semantically describe the solution. The paper shows empirically that AGE is as good as GE for a classical problem, and proves that including semantics in the grammar can improve GE performance. An important conclusion is that adding too much semantics can make the search difficult.

We present a new classification system based on Evolutionary Algorithm (EA), OBLIC. This tool is an OBLIque Classification system whose function is to induce a set of classification rules no hierarchical from a database or training set. The core of the algorithm is a EA with real-coded and Pittsburgh approach. Each individual is composed by a no fixed classification rules set what split in regions the search space. The fitness of each classification is obtained by means of the exploration of these regions. The result of the tool is the best classification obtained in the evolutionary process.This paper describe and analyze this new method by comparing with other classification systems on UCI Repository databases. We conclude this paper with some observations and future projects.

International audience; This paper introduces a new kind of skeleton for binary volumes called the cellular skeleton. This skeleton is not a subset of voxels of a volume nor a subcomplex of a cubical complex: it is a chain complex together with a reduction from the original complex.Starting from the binary volume we build a cubical complex which represents it regarding 6 or 26-connectivity. Then the complex is thinned using the proposed method based on elementary collapses, which preserves significant geometric features. The final step reduces the number of cells using Discrete Morse Theory. The resulting skeleton is a reduction which preserves the homology of the original complex and the geometrical information of the output of the previous step.The result of this method, besides its skeletonization content, can be used for computing the homology of the original complex, which usually provides well shaped homology generators.

4th International Conference, SIMPAR 2014, Bergamo, Italy, October 20-23, 2014, Field experiments with a team of heterogeneous robots require human and hardware resources which cannot be implemented in a straightforward manner. Therefore, simulation environments are viewed by the robotic community as a powerful tool that can be used as an intermediate step to evaluate and validate the developments prior to their integration in real robots. This paper evaluates a novel multi-robot heterogeneous cooperative perception framework based on monocular measurements under the MORSE robotic simulation environment. The simulations are performed in an outdoor environment using a team of Micro Aerial Vehicles (MAV) and an Unmanned Ground Vehicle (UGV) performing distributed cooperative perception based on monocular measurements. The goal is to estimate the 3D target position.

Date du colloque: 09/2012; International audience; The max-cut problem is to partition the vertices of a weighted graph G = (V,E) into two subsets such that the weight sum of the edges crossing the two subsets is maximized. This paper presents a memetic max-cut algorithm (MACUT) that relies on a dedicated multi-parent crossover operator and a perturbation-based tabu search procedure. Experiments on 30 G-set benchmark instances show that MACUT competes favorably with 6 state-of-the-art max-cut algorithms, and for 10 instances improves on the best known results ever reported in the literature.

This paper proposes a Radial Basis Function Neural Network (RBFNN) which reproduces different Radial Basis Functions (RBFs) by means of a real parameter q, named q-Gaussian RBFNN. The architecture, weights and node topology are learnt through a Hybrid Algorithm (HA) with the iRprop + algorithm as the local improvement procedure. In order to test its overall performance, an experimental study with four gene microarray datasets with two classes taken from bioinformatic and biomedical domains is presented. The Fast Correlation–Based Filter (FCBF) was applied in order to identify salient expression genes from thousands of genes in microarray data that can directly contribute to determining the class membership of each pattern. After different gene subsets were obtained, the proposed methodology was performed using the selected gene subsets as the new input variables. The results confirm that the q-Gaussian RBFNN classifier leads to promising improvement on accuracy.

Land use and land cover (LULC) maps are remote sensing products that are used to classify areas into different landscapes. The newest techniques have been applied to improve the final LULC classification and most of them are based on SVM classifiers. In this paper, a new method based on a multiple classifiers ensemble to improve LULC map accuracy is shown. The method builds a statistical raster from LIDAR and image fusion data following a pixel-oriented strategy. Then, the pixels from a training area are used to build a SVM and k-NN restricted stacking taking into account the special characteristics of spatial data. A comparison between a SVM and the restricted stacking is carried out. The results of the tests show that our approach improves the results in the context of the real data from a riparian area of Huelva (Spain).

Date du colloque: 04/2009; International audience; The Maximum Parsimony problem aims at reconstructing a phylogenetic tree from DNA, RNA or protein sequences while minimizing the number of evolutionary changes. Much work has been devoted by the research community to solve this NP-complete problem and many algorithms and techniques have been devised in order to find high quality solutions with reasonable computational resources. In this paper we present a memetic algorithm (implemented in the software Hydra) which is based on an integration of an effective local search operator with a specific topological tree crossover operator. We report computational results of Hydra on a set of 12 benchmark instances from the literature and demonstrate its effectiveness with respect to one of the most powerful software (TNT). We also study the behavior of the algorithm with respect to some fundamental ingredients.

The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-642-03741-2_26, Proceedings of Third International Conference, CALCO 2009, Udine, Italy, September 7-10, 2009., Model-to-model (M2M) transformation consists in trans- forming models from a source to a target language. Many transformation languages exist, but few of them combine a declarative and relational style with a formal underpinning able to show properties of the transformation. Pattern-based transformation is an algebraic, bidirectional, and relational approach to M2M transformation. Specifications are made of patterns stating the allowed or forbidden relations between source and target models, and then compiled into low level operational mechanisms to perform source-to-target or target-to-source transformations. In this paper, we study the compilation into operational triple graph grammar rules and show: (i) correctness of the compilation of a specification without negative patterns; (ii) termination of the rules, and (iii) completeness, in the sense that every model considered relevant can be built by the rules., Work supported by the Spanish Ministry of Science and Innovation, projects METEORIC (TIN2008-02081), MODUWEB (TIN2006-09678) and FORMALISM (TIN2007-66523). Moreover, part of this work was done during a sabbatical leave of the first author at TU Berlin, with financial support from the Spanish Ministry of Science and Innovation (grant ref. PR2008-0185). We thank the referees for their useful comments

Date du colloque: 04/2009; International audience; In this paper, we study the perturbation operator of Iterated Local Search. To guide more efficiently the search to move towards new promising regions of the search space, we introduce a Critical Element-Guided Perturbation strategy (CEGP). This perturbation approach consists of the identification of critical elements and then focusing on these critical elements within the perturbation operator. Computational experiments on two case studies—graph coloring and course timetabling—give evidence that this critical element-guided perturbation strategy helps reinforce the performance of Iterated Local Search.

The final publication is available at Springer via http://dx.doi.org/10.1007/978-3-540-87405-8_29, Proceedings of 4th International Conference, ICGT 2008, Leicester, United Kingdom, September 7-13, 2008, We present a new, high-level approach for the specification of model-to-model transformations based on declarative patterns. These are (atomic or composite) constraints on triple graphs declaring the allowed or forbidden relationships between source and target models. In this way, a transformation is defined by specifying a set of triple graph constraints that should be satisfied by the result of the transformation. The description of the transformation is then compiled into lower-level operational mechanisms to perform forward or backward transformations, as well as to establish mappings between two existent models. In this paper we study one of such mechanisms based on the generation of operational triple graph grammar rules. Moreover, we exploit deduction techniques at the specification level in order to generate more specialized constraints (preserving the specification semantics) reflecting pattern dependencies, from which additional rules can be derived., Work supported by the Spanish Ministry of Education and Science, projects TSI2005-08225-C07-06 and TIN2006-09678.

The final publication is available at Springer via http://dx.doi.org/10.1007/11841883_25, Proceedings of Third International Conference, ICGT 2006 Natal, Rio Grande do Norte, Brazil, September 17-23, 2006, In this paper, we present our approach for model view management in the context of Multi-View Visual Languages (MVVLs). These are made of a number of diagram types (or viewpoints) that can be used for the specification of the different aspects of a system. Therefore, the user can build different system views conform to the viewpoints, which are merged in a repository in order to perform consistency checking. In addition, the user can define derived views by means of graph query patterns in order to extract information from a base model (a system view or the repository). We have provided automatic mechanisms to keep synchronized the base model and the derived view when the former changes. Predefined queries by the MVVL designer result in so-called audience-oriented views. Finally, semantic views are used for analysing the system by its translation into a semantic domain. Our approach is based on meta-modelling to describe the syntax of the MVVL and each viewpoint, and on triple graph transformation systems to synchronize and maintain correspondences between the system views and the repository, as well as between the derived, audience-oriented and semantic views and the base models. We illustrate these concepts by means of an example in the domain of security for web systems., This work has been sponsored by the Spanish Ministry of Science and Education, projects TSI2005-08225-C07-06 and TSI2004-03394.

The final publication is available at Springer via http://dx.doi.org/10.1007/11840817_25, Proceedings of 16th International Conference on Artificial Neural Networks, Athens, Greece, September 10-14, 2006, Part I, Recently, bump formations in attractor neural networks with distance dependent connectivities has become of increasing interest for investigation in the field of biological and computational neuroscience. Although the distance dependent connectivity is common in biological networks, a common fault of these network is the sharp drop of the number of patterns p that can remembered, when the activity changes from global to bump-like, than effectively makes these networks low effective. In this paper we represent a bump-based recursive network specially designed in order to increase its capacity, which is comparable with that of randomly connected sparse network. To this aim, we have tested a selection of 700 natural images on a network with N = 64K neurons with connectivity per neuron C. We have shown that the capacity of the network is of order of C, that is in accordance with the capacity of highly diluted network. Preserving the number of connections per neuron, a non-trivial behavior with the radius of the connectivity has been observed. Our results show that the decrement of the capacity of the bumpy network can be avoided., The authors acknowledge the financial support from the Spanish Grants DGI.M. CyT. FIS2005-1729, Plan de Promoción de la Investigación UNED and TIN 2004–07676-G01-01.We also thank David Dominguez for the fruitful discussion of the manuscript.

The final publication is available at Springer via http://dx.doi.org/10.1007/11550822_78, Proceedings of 15th International Conference, Warsaw, Poland, September 11-15, 2005., Model transformation is one of the pillars of Model-Driven Engineering (MDE). The increasing complexity of systems and modelling languages has dramatically raised the complexity and size of model transformations. Even though many transformation languages and tools have been proposed in the last few years, most of them are directed to the implementation phase of transformation development. However, there is a lack of cohesive support for the other phases of the transformation development, like requirements, analysis, design and testing. In this paper, we propose a unified family of languages to cover the life-cycle of transformation development. Moreover, following an MDE approach, we provide tools to partially automate the progressive refinement of models between the different phases and the generation of code for specific transformation implementation languages., This work has been partially supported by grants BFI2003-07276 and TIN2004-04363-C03-03, and by PROFIT project FIT-330100-2004-91

Proceedings of Joint International Conference ICANN/ICONIP 2003 Istanbul, Turkey, June 26–29, 2003, The final publication is available at Springer via http://dx.doi.org/10.1007/3-540-44989-2_50, Image superresolution methods process an input image sequence of a scene to obtain a still image with increased resolution. Classical approaches to this problem involve complex iterative minimization procedures, typically with high computational costs. In this paper is proposed a novel algorithm for superresolution that enables a substantial decrease in computer load. First, a probabilistic neural network architecture is used to perform a scattered-point interpolation of the image sequence data. The network kernel function is optimally determined for this problem by a multi-layer perceptron trained on synthetic data. Network parameters dependence on sequence noise level is quantitatively analyzed. This super-sampled image is spatially filtered to correct finite pixel size effects, to yield the final high-resolution estimate. Results on a real outdoor sequence are presented, showing the quality of the proposed method., This work has been partially supported by TIC2001-0572-C02-02 grant

E–mail is one of the most common ways to communicate, assuming, in some cases, up to 75% of a company’s communication, in which every employee spends about 90 minutes a day in e–mail tasks such as filing and deleting. This paper deals with the generation of clusters of relevant words from E–mail texts. Our approach consists of the application of text mining techniques and, later, data mining techniques, to obtain related concepts extracted from sent and received messages. We have developed a new clustering algorithm based on neighborhood, which takes into account similarity values among words obtained in the text mining phase. The potential of these applications is enormous and only a few companies, mainly large organizations, have invested in this project so far, taking advantage of employees’s knowledge in future decisions.

This paper describes a time-series prediction method based on the kNN technique. The proposed methodology is applied to the 24-hour load forecasting problem. Also, based on recorded data, an alternative model is developed by means of a conventional dynamic regression technique, where the parameters are estimated by solving a least squares problem. Finally, results obtained from the application of both techniques to the Spanish transmission system are compared in terms of maximum, average and minimum forecasting errors.

The final publication is available at Springer via http://dx.doi.org/10.1007/3-540-46084-5_5, Proceedings of International Conference Madrid, Spain, August 28–30, 2002, Small-World networks are highly clusterized networks with small distances between their nodes. There are some well known biological networks that present this kind of connectivity. On the other hand, the usual models of Small-World networks make use of undirected and unweighted graphs in order to represent the connectivity between the nodes of the network. These kind of graphs cannot model some essential characteristics of neural networks as, for example, the direction or the weight of the synaptic connections. In this paper we analyze different kinds of directed graphs and show that they can also present a Small-World topology when they are shifted from regular to random. Also analytical expressions are given for the cluster coefficient and the characteristic path of these graphs., We thank the Ministerio de Ciencia y Tecnología (BFI 2000-015). (RH) was also funded by DE-FG03-96ER14092, (CA) was partially supported by ARO-MURI grant DAA655-98-1-0249 during a four month stay in UCSD. (PP) and (CA) are partially supported by PB98-0850

Graphical features on map, charts, diagrams and graph drawings usually must be annotated with text labels in order to convey their meaning. In this paper we focus on a problem that arises when labeling schematized maps, e.g. for subway networks. We present algorithms for labeling points on a line with axis-parallel rectangular labels of equal height. Our aim is to maximize label size under the constraint that all points must be labeled. Even a seemingly strong simplification of the general point-labeling problem, namely to decide whether a set of points on a horizontal line can be labeled with sliding rectangular labels, turns out to be weakly NPcomplete. This is the first labeling problem that is known to belong to this class. We give a pseudo-polynomial time algorithm for it. In case of a sloping line points can be labeled with maximum-size square labels in O(n log n) time if four label positions per point are allowed and in O(n 3 log n) time if labels can slide. We also investigate rectangular labels.