Benchmarking Least Squares Support Vector Machine Classifiers

In Support Vector Machines (SVMs), the solution of the classification problem is characterized by a (convex) quadratic programming (QP) problem. In a modified version of SVMs, called Least Squares SVM classifiers (LS-SVMs), a least squares cost function is proposed so as to obtain a linear set of equations in the dual space. While the SVM classifier has a large margin interpretation, the LS-SVM formulation is related in this paper to a ridge regression approach for classification with binary targets and to Fisher's linear discriminant analysis in the feature space. Multiclass categorization problems are represented by a set of binary classifiers using different output coding schemes. While regularization is used to control the effective number of parameters of the LS-SVM classifier, the sparseness property of SVMs is lost due to the choice of the 2-norm. Sparseness can be imposed in a second stage by gradually pruning the support value spectrum and optimizing the hyperparameters during the sparse approximation procedure. In this paper, twenty public domain benchmark datasets are used to evaluate the test set performance of LS-SVM classifiers with linear, polynomial and radial basis function (RBF) kernels. Both the SVM and LS-SVM classifier with RBF kernel in combination with standard cross-validation procedures for hyperparameter selection achieve comparable test set performances. These SVM and LS-SVM performances are consistently very good when compared to a variety of methods described in the literature including decision tree based algorithms, statistical algorithms and instance based learning methods. We show on ten UCI datasets that the LS-SVM sparse approximation procedure can be successfully applied.     

Future train traffic control: control by re-planning

Improving train traffic control can be a cost-efficient way to improve train traffic punctuality and increase utilization of existing and future railway infrastructure. However, performance in train traffic control tasks currently involves working on a technical level in order to regulate the traffic flow. Working in a preventive manner is poorly supported and train traffic controllers are usually restricted to just solving problems as they occur. This often results in unnecessarily long delays and decreased timeliness of train traffic. The main objective of this paper is to describe a proposed control strategy and a case study, which evaluates the control strategy and the prototype tool derived from the research. By shifting the control paradigm to a high-level control strategy, many of today’s problems may be avoided, with benefits of the reduction in delays, improved timeliness and better utilization of the infrastructure. Twenty-one train traffic controllers participated in a case study, with a simulated prototype environment. The majority of the participating train traffic controllers were positive to the new concepts and ideas. Many of the important aspects of the proposed control strategy can be investigated with the simulation, but due to the complexity of train traffic some issues must be evaluated in an operative environment.     

Recombination luminescence from defects in boron-ion implantation-doped diamond using low fluences

 Two identical, high purity, natural type IIa diamonds, which displayed the ubiquitous blue cathodoluminescence (CL) band at ≈ 2.9 eV, as well as an indication of the corresponding green band at ≈ 2.4 eV, have been equivalently doped by using extremely low dose B⁺- and C⁺-ion CIRA-implantations respectively. Comparative CL measurements showed changes in the intensities of the 2.9 and 2.4 eV bands and the generation of bands at ≈ 4 eV, as well as at ≈ 3.5 and ≈ 4.6 eV (the latter two in the B⁺-CIRA diamond). The results are commensurate with the model (proposed previously) in which the 2.9 and 4 eV bands are generated respectively by electron-hole recombinations at negatively charged acceptor- and positively charged donor-like, intrinsic defects. The present results indicate that Coulomb interactions between the latter defects and (at least partially) compensated, negatively charged, boron acceptors, generate the 3.5 and 4.6 eV bands, which may be considered as higher energy (≈ 0.6 eV) replicas of the 2.9 and 4 eV bands. In both cases, two electrons and a hole interact just before the hole combines with an electron. Such a configuration of charges seems related to, and could possibly be described as, a type of ”ionised exciton molecule”, where the ”bonding” of two negative ”nuclei” is facilitated by the presence of the hole. The CL measurements further indicate that the 2.4 eV band forms when a high enough density of, in this case, neutral acceptors are present. These neutral acceptors compete with the valence band to supply holes for recombination at the negatively charged, acceptor-type, intrinsic defects which are, in the absence of the boron, responsible for the generation of the blue, 2.9 eV band. Received: 5 December 1997 / Accepted: 13 December 1997     

Load forecasting using a multivariate meta-learning system

Although over a thousand scientific papers address the topic of load forecasting every year, only a few are dedicated to finding a general framework for load forecasting that improves the performance, without depending on the unique characteristics of a certain task such as geographical location. Meta-learning, a powerful approach for algorithm selection has so far been demonstrated only on univariate time-series forecasting. Multivariate time-series forecasting is known to have better performance in load forecasting. In this paper we propose a meta-learning system for multivariate time-series forecasting as a general framework for load forecasting model selection. We show that a meta-learning system built on 65 load forecasting tasks returns lower forecasting error than 10 well-known forecasting algorithms on 4 load forecasting tasks for a recurrent real-life simulation. We introduce new metafeatures of fickleness, traversity, granularity and highest ACF. The meta-learning framework is parallelized, component-based and easily extendable.     

Development of a Modelica Compiler Using JastAdd

This paper describes experiences from implementing key parts of a compiler for Modelica, an object-oriented language supporting declarative modeling and simulation of complex physical systems. Our implementation uses the attribute-grammar based tool JastAdd. In particular, we discuss the implementation of Modelica name analysis which is highly context-dependent, type analysis which is based on structural subtyping, and flattening which is a fundamental part of the Modelica compilation process.of so called modifications, Modelica.     

Enforcing structural regularities in software using IntensiVE

The design and implementation of a software system is often governed by a variety of coding conventions, design patterns, architectural guidelines, design rules, and other so-called structural regularities. To prevent a deterioration of the system’s source code, it is important that these regularities are verified and enforced upon evolution of the system. The Intensional Views Environment (IntensiVE), presented in this article, is a tool suite for specifying relevant structural regularities in an (object-oriented) software system and verifying them against the current and later versions of the system. At the heart of the IntensiVE tool suite are (logic) program queries and the model of intensional views and relations, through which regularities are expressed. Upon verification of these regularities in the source code of the system, IntensiVE reports the code entities (i.e. classes, methods, variables, statements, etc.) that violate these constraints. We present IntensiVE and illustrate its application to the verification of an Abstract Factory design pattern in the implementation of a software system.     

The GOEDEL Program

The GOEDEL program is an ever-growing collection of currently more than thirty thousand rewrite rules for transforming expressions in Gödel’s class theory in the hope of thereby simplifying them. A brief survey of the program and its use is presented. A few striking results obtained recently using the program are featured to provide some of the flavor of the author’s on-going research over the past fifteen years.     

W-based versus latent variables spatial autoregressive models: evidence from Monte Carlo simulations

In this paper, we compare by means of Monte Carlo simulations two approaches to take spatial autocorrelation into account: the classical spatial autoregressive model and the structural equations model with latent variables. The former accounts for spatial dependence and spillover effects in georeferenced data by means of a spatial weights matrix W. The latter represents spatial dependence and spillover effects by means of a latent variable in the structural (regression) model while the observed spatially lagged variables are related to the latent spatial dependence variable in the measurement model. The simulation results based on Anselin’s Columbus, Ohio, crime data set show that the misspecified latent variables approach slightly trails the correctly specified classical approach in terms of bias and root mean squared error of the coefficient estimators.     

Dynamic prototypicality effects in visual search.

In recent studies, researchers have discovered a larger neural activation for stimuli that are more extreme exemplars of their stimulus class, compared with stimuli that are more prototypical. This has been shown for faces as well as for familiar and novel shape classes. We used a visual search task to look for a behavioral correlate of these findings regarding both simple geometrical shapes and more complex, novel shape classes. The latter stimulus set enabled us to control for the physical properties of the shapes, establishing that the effects are solely due to the positions of the particular stimuli in a particular shape space (i.e., more extreme versus more central in shape space) and not to specific shape features. The results indicate that finding an atypical instance of a shape class among more prototypical ones is easier and faster than the other way around. The prototypical status of a shape in our experiment could change very quickly, that is, within minutes, depending on the subset of shapes that was shown to the participants. Manipulating the degree of familiarity toward the shapes by selectively increasing familiarity for the extreme shapes did not influence our results. In general, we show that the prototypical status of a stimulus in visual search is a highly dynamic property, depending on the distribution of stimuli within a shape space but not on familiarity with the prototype. (PsycINFO Database Record (c) 2011 APA, all rights reserved)     

How can you resist? Executive control helps romantically involved individuals to stay faithful.

In the present research, we examined why some people have more difficulty than others in staying faithful to their romantic partners. Three studies supported our main prediction that executive control is associated with romantically involved individuals' ability to stay faithful. Study 1 showed that participants with a higher level of executive control reported less difficulty in staying faithful to their partners than did those with lower levels of executive control. In Study 2, romantically involved male participants were placed in a waiting room together with an attractive female confederate. Results showed that participants with a higher level of executive control showed less flirting behavior with the confederate than did those with lower levels of executive control. Study 3 demonstrated that a higher level of executive control was related to a lower expressed desire to meet an attractive other, but only for romantically involved participants. Together, these studies showed that executive control helps romantically involved individuals to deal with the lure of attractive alternatives. (PsycINFO Database Record (c) 2011 APA, all rights reserved)