Collaborative filtering with interlaced generalized linear models

Collaborative filtering (CF) is a data analysis task appearing in many challenging applications, in particular data mining in Internet and e-commerce. CF can often be formulated as identifying patterns in a large and mostly empty rating matrix. In this paper, we focus on predicting unobserved ratings. This task is often a part of a recommendation procedure. We propose a new CF approach called interlaced generalized linear models (GLM); it is based on a factorization of the rating matrix and uses probabilistic modeling to represent uncertainty in the ratings. The advantage of this approach is that different configurations, encoding different intuitions about the rating process can easily be tested while keeping the same learning procedure. The GLM formulation is the keystone to derive an efficient learning procedure, applicable to large datasets. We illustrate the technique on three public domain datasets.     

Configuration for mass customization: how to extend product configuration towards requirements and process configuration

In order to develop mass customization, many companies use configuration software to customize their products. Although many studies already exist about Product Configuration, Requirements and Process Configuration have not been studied in detail. As all these three aspects must be considered for mass customization, the aim of this paper is to show how Product Configuration, when considered as a constraint satisfaction problem, can be extended upstream towards Requirements Configuration and downstream towards Process Configuration. Product Configuration basics are first reviewed thanks to a constraint based approach, and an analysis of industrial configuration situations is done in order to clarify mass customization needs in terms of configuration. Then upstream Requirements Configuration and downstream Process Configuration are defined and generic models are proposed. It is shown that the proposed elements allow a global and consistent flow of configuration activities. A detailed example illustrates the different configuration problems and a discussion terminates the paper.     

Phase-sensitive spatially-modulated surface plasmon resonance polarimetry for detection of biomolecular interactions

We here describe a scheme of spatially modulated surface plasmon resonance (SPR) polarimetry that enables to combine ultra-high phase sensitivity with good signal-to-noise background. The proposed approach uses spatial modulation of s-polarized component by birefringent elements and the extraction of phase-polarization information by Fourier-transform methods. This scheme was tested for monitoring the interactions between an antibody and its biological partner. Our experimental data, collected by amplitude-sensitive and phase-sensitive polarimetry demonstrate that the latter scheme provides at least one order of magnitude improvement in terms of detection limit.     

Model-driven visualizations of constraint-based local search

Visualization is often invaluable to understand the behavior of optimization algorithms, identify their bottlenecks or pathological behaviors, and suggest remedial techniques. Yet developing visualizations is often a tedious activity requiring significant time and expertise. This paper presents a framework for the visualization of constraint-based local search (CBLS) algorithms. Given a high-level model and a declarative visualization specification, the CBLS visualizer systematically produces animations to visualize constraints and objectives, violations, and conflicts, as well as the temporal behavior of these measures. The visualization specification is declarative and typically composed of a triple (what,where,how) indicating what to display, where, and with which graphical objects. The visualizer architecture is compositional and extensible. It provides building blocks which can be assembled freely by the user and focuses almost exclusively on static aspects, the dynamic aspects being automated by the use of invariants. The paper highlights various functionalities of the visualizer and describes a blueprint for its implementation.     

A kinetic interpretation of a negative time constant in impedance equivalent circuits for the dissolution/passive transition

A theoretical impedance function is proposed for the active/passive transition of nickel in an acid sulphate medium. It is considered that passivating species progressively covers the electrode surface. This approximated model predicts the appearance of negative time constants in the impedance spectra when coverage coefficient θ values for passive species are greater than 0.5.     

Statistiques de la fonction de transfert par trajets multiples pour des largeurs de bande moyennes. Application à la prévision de la qualité

It is shown, from data gathered on three different test links, that multipath transfer functions on a medium bandwidth can be modelled with three parameters. Moreover, the model parameters follow similar probability laws on the three links, thus leading to a universalstatistical model. This statistical model depends on two independent parameters. When the two statistical parameters are known, the outage time of the link can be deduced from the model. A first estimation of these parameters can be inferred from the distribution law of the levels at fixed frequency. The application of the method to two particular cases is presented.     

A survey of ellagitannin content in raspberry and blackberry cultivars grown in Trentino (Italy)

This study describes the first application of a new procedure for acid hydrolysis of Rubus ellagitannins in methanol, which allows quantification of all the major reaction products (ellagic acid, methyl-sanguisorboate, methyl gallate and an unknown ellagic acid derivative) and estimation of the mean degree of polymerization (mDP) of Rubus ellagitannins. The ellagitannins of 21 raspberry and 15 blackberry cultivars produced in Trentino, Italy during 2005 were extracted and characterised. The total content of ellagitannins was shown to vary by a factor of 2 among raspberry cultivars and by a factor of 4 among blackberry cultivars.     

Glucose and Fructose Fermentation by Wine Yeasts in Media Containing Structurally Complex Nitrogen Sources

Glucose and fructose fermentations by industrial yeasts strains are strongly affected by both the structural complexity of the nitrogen source and the availability of oxygen. In this study two Saccharomyces cerevisiae industrial wine strains were grown, under shaken and static conditions, in a media containing either a) 20% (w/v) glucose, or b) 10% (w/v) fructose and 10% (w/v) glucose or c) 20% (w/v) fructose, all supplemented with nitrogen sources varying from a single ammonium salt (ammonium sulfate) to free amino acids (casamino acids) and peptides (peptone). Data suggest that a complex structured nitrogen source is not submitted to the same control mechanisms as those involved in the utilization of simpler structured nitrogen sources, and mutual interaction between carbon and nitrogen sources, including the mechanisms involved in the regulation of aerobic/anaerobic metabolism, may play an important role in defining yeast fermentation performance and the differing response to the structural complexity of the nitrogen source, with a strong impact on fermentation performance.     

A nuclear reactor core fuel reload optimization using artificial ant colony connective networks

The core of a nuclear Pressurized Water Reactor (PWR) may be reloaded every time the fuel burn-up is such that it is not more possible to maintain the reactor operating at nominal power. The nuclear core fuel reload optimization problem consists in finding a pattern of burned-up and fresh-fuel assemblies that maximize the number of full operational days. This is an NP-Hard problem, meaning that complexity grows exponentially with the number of fuel assemblies in the core. Moreover, the problem is non-linear and its search space is highly discontinuous and multi-modal.