New kernel functions and learning methods for text and data mining

Recent advances in machine learning methods enable increasingly the automatic construction of various types of computer assisted methods that have been difficult or laborious to program by human experts. The tasks for which this kind of tools are needed arise in many areas, here especially in the fields of bioinformatics and natural language processing. The machine learning methods may not work satisfactorily if they are not appropriately tailored to the task in question. However, their learning performance can often be improved by taking advantage of deeper insight of the application domain or the learning problem at hand. This thesis considers developing kernel-based learning algorithms incorporating this kind of prior knowledge of the task in question in an advantageous way. Moreover, computationally efficient algorithms for training the learning machines for specific tasks are presented. In the context of kernel-based learning methods, the incorporation of prior knowledge is often done by designing appropriate kernel functions. Another well-known way is to develop cost functions that fit to the task under consideration. For disambiguation tasks in natural language, we develop kernel functions that take account of the positional information and the mutual similarities of words. It is shown that the use of this information significantly improves the disambiguation performance of the learning machine. Further, we design a new cost function that is better suitable for the task of information retrieval and for more general ranking problems than the cost functions designed for regression and classification. We also consider other applications of the kernel-based learning algorithms such as text categorization, and pattern recognition in differential display. We develop computationally efficient algorithms for training the considered learning machines with the proposed kernel functions. We also design a fast cross-validation algorithm for regularized least-squares type of learning algorithm. Further, an efficient version of the regularized least-squares algorithm that can be used together with the new cost function for preference learning and ranking tasks is proposed. In summary, we demonstrate that the incorporation of prior knowledge is possible and beneficial, and novel advanced kernels and cost functions can be used in algorithms efficiently.

Capability matrix – Identifying and evaluating the key capabilities of purchasing and supply management

Organizations gain resources, skills and technologies to find out the ultimate mix of capabilities to be a winner in the competitive market. These are all important factors that need to be taken into account in organizations operating in today's business environment. So far, there are no significant studies on the organizational capabilities in the field of PSM. The literature review shows that the PSM capabilities need to be studied more comprehensively. This study attempts to reveal and fill this gap by providing the PSM capability matrix that identifies the key PSM capabilities approached from two angles: there are three primary PSM capabilities and nine subcapabilities and, moreover, the individual and organizational PSM capabilities are identified and evaluated. The former refers to the PSM capability matrix of this study which is based on the strategic and operative PSM capabilities that complement the economic ones, while the latter relates to the evaluation of the PSM capabilities, such as the buyer profiles of individual PSM capabilities and the PSMcapability map of the organizational ones. This is a constructive case study. The aim is to define what the purchasing and supply management capabilities are and how they can be evaluated. This study presents a PSM capability matrix to identify and evaluate the capabilities to define capability gaps by comparing the ideal level of PSM capabilities to the realized ones. The research questions are investigated with two case organizations. This study argues that PSM capabilities can be classified into three primary categories with nine sub-categories and, thus, a PSM capability matrix with four evaluation categories can be formed. The buyer profiles are moreover identified to reveal the PSM capability gap. The resource-based view (RBV) and dynamic capabilities view (DCV) are used to define the individual and organizational capabilities. The PSM literature is also used to define the capabilities. The key findings of this study are i) the PSM capability matrix to identify the PSM capabilities, ii) the evaluation of the capabilities to define PSM capability gaps and iii) the presentation of the buyer profiles to identify the individual PSM capabilities and to define the organizational PSM capabilities. Dynamic capabilities are also related to the PSM capability gap. If a gap is identified, the organization can renew their PSM capabilities and, thus, create mutual learning and increase their organizational capabilities. And only then, there is potential for dynamic capabilities. Based on this, the purchasing strategy, purchasing policy and procedures should be identified and implemented dynamically.

What can ecosystems learn? Expanding evolutionary ecology with learning theory.

BACKGROUND: The structure and organisation of ecological interactions within an ecosystem is modified by the evolution and coevolution of the individual species it contains. Understanding how historical conditions have shaped this architecture is vital for understanding system responses to change at scales from the microbial upwards. However, in the absence of a group selection process, the collective behaviours and ecosystem functions exhibited by the whole community cannot be organised or adapted in a Darwinian sense. A long-standing open question thus persists: Are there alternative organising principles that enable us to understand and predict how the coevolution of the component species creates and maintains complex collective behaviours exhibited by the ecosystem as a whole? RESULTS: Here we answer this question by incorporating principles from connectionist learning, a previously unrelated discipline already using well-developed theories on how emergent behaviours arise in simple networks. Specifically, we show conditions where natural selection on ecological interactions is functionally equivalent to a simple type of connectionist learning, 'unsupervised learning', well-known in neural-network models of cognitive systems to produce many non-trivial collective behaviours. Accordingly, we find that a community can self-organise in a well-defined and non-trivial sense without selection at the community level; its organisation can be conditioned by past experience in the same sense as connectionist learning models habituate to stimuli. This conditioning drives the community to form a distributed ecological memory of multiple past states, causing the community to: a) converge to these states from any random initial composition; b) accurately restore historical compositions from small fragments; c) recover a state composition following disturbance; and d) to correctly classify ambiguous initial compositions according to their similarity to learned compositions. We examine how the formation of alternative stable states alters the community's response to changing environmental forcing, and we identify conditions under which the ecosystem exhibits hysteresis with potential for catastrophic regime shifts. CONCLUSIONS: This work highlights the potential of connectionist theory to expand our understanding of evo-eco dynamics and collective ecological behaviours. Within this framework we find that, despite not being a Darwinian unit, ecological communities can behave like connectionist learning systems, creating internal conditions that habituate to past environmental conditions and actively recalling those conditions. REVIEWERS: This article was reviewed by Prof. Ricard V Solé, Universitat Pompeu Fabra, Barcelona and Prof. Rob Knight, University of Colorado, Boulder.

Customer-Related Internal Communication in a Global Matrix Organization

The objective of the thesis was to explore the nature and characteristics of customer-related internal communication in a global industrial matrix organization during a specific customer relationship, and how it could be improved. The theoretical part of the study views the field of the concepts of intra-organizational information and knowledge sharing. The theoretical part also views the internal communications influences to customer relationships, its problematic, and the suggestions to improve internal communication in literature. The empirical part of the study was conducted with the Content Analysis and the Social Network Analysis as research methods. The data was collected by interviews and a questionnaire. Internal communication was observed first generally within the organization from the point of view of a certain business, and secondly, during a specific customer relationship at personal level and at departmental level. The results of the study describe the nature and characteristics of internal communication in the organization. The results give 13 suggestions for improving internal communication in the organization. Although the study has been done in one specific organization, it also offers insights for other organizations as well as managers to improve their internal communication.

New insights into cellular prion protein (PrPc) functions: the 'ying and yang' of a relevant protein.

The conversion of cellular prion protein (PrPc), a GPI-anchored protein, into a protease-K-resistant and infective form (generally termed PrPsc) is mainly responsible for Transmissible Spongiform Encephalopathies (TSEs), characterized by neuronal degeneration and progressive loss of basic brain functions. Although PrPc is expressed by a wide range of tissues throughout the body, the complete repertoire of its functions has not been fully determined. Recent studies have confirmed its participation in basic physiological processes such as cell proliferation and the regulation of cellular homeostasis. Other studies indicate that PrPc interacts with several molecules to activate signaling cascades with a high number of cellular effects. To determine PrPc functions, transgenic mouse models have been generated in the last decade. In particular, mice lacking specific domains of the PrPc protein have revealed the contribution of these domains to neurodegenerative processes. A dual role of PrPc has been shown, since most authors report protective roles for this protein while others describe pro-apoptotic functions. In this review, we summarize new findings on PrPc functions, especially those related to neural degeneration and cell signaling.

Two-year follow-op of cognitive functions in Schizophrenia spectrum disorders of adolescent patients treated with electroconvulsive therapy

Objective: The aim of the current study was to investigate the long-term cognitive effects of electroconvulsive therapy (ECT) in a sample of adolescent patients in whom schizophrenia spectrum disorders were diagnosed. Methods: The sample was composed of nine adolescent subjects in whom schizophrenia or schizoaffective disorder was diagnosed according to DSM-IV-TR criteria on whom ECT was conducted (ECT group) and nine adolescent subjects matched by age, socioeconomic status, and diagnostic and Positive and Negative Syndrome Scale (PANSS) total score at baseline on whom ECT was not conducted (NECT group). Clinical and neuropsychological assessments were carried out at baseline before ECT treatment and at 2-year follow-up. Results: Significant differences were found between groups in the number of unsuccessful medication trials. No statistically significant differences were found between the ECT group and theNECT group in either severity as assessed by the PANSS, or in any cognitive variables at baseline.At follow-up, both groups showed significant improvement in clinical variables (subscales of positive, general, and total scores of PANSS and Clinical Global Impressions-Improvement). In the cognitive assessment at follow-up, significant improvement was found in both groups in the semantic category of verbal fluency task and digits forward. However, no significant differences were found between groups in any clinical or cognitive variable at follow-up. Repeated measures analysis found no significant interaction of time · group in any clinical or neuropsychological measures. Conclusions: The current study showed no significant differences in change over time in clinical or neuropsychological variables between the ECT group and the NECT group at 2-year follow-up. Thus, ECT did not show any negative influence on long-term neuropsychological variables in our sample.

Inactive matrix gla-protein is associated with arterial atiffness in an adult population-based study

La vitesse de l'onde de pouls (VOP) est la méthode pour mesurer la rigidité artérielle la plus répandue et la plus validée. C'est aussi un prédicteur indépendant de la mortalité. La Matrix Gla- protein (MGP) est une protein qui inhibe les calcifications vasculaires. MGP nécessite une enzyme dérivée de la vitamine K pour être activée, à l'instar de certains facteurs de coagulation. La forme inactive de MGP, connue sous le terme de « desphospho-uncarboxylated MGP » (dp-ucMGP), peut-être mesurée dans le plasma. Plus les apports de vitamine K sont importants plus les taux de dp-ucMGP diminue. Les taux de dp-ucMGP ont déjà été étudiés et associés à différents marqueurs cardiovasculaires (CV), aux événements CV et à la mortalité. Dans notre travail de recherche nous avons émis l'hypothèse que des taux élevés de dp-ucMGP seraient associés à une VOP élevée. Nous avons recruté les participants à travers une étude multicentrique suisse (SKIPOGH). Le processus de recrutement ciblait des familles dans lesquelles plusieurs membres étaient d'accord de participer. Nous avons mesuré la dp-ucMGP plasmatique grâce à la méthode immuno-enzymatique « ELISA ». Concernant la VOP, nous avons mesuré les ondes de pression au niveau carotidien et fémorale grâce à un tonomètre et calculer la vitesse de leurs propagations. Par la suite nous avons utilisé un modèle de régression linéaire multiple afin de déterminer le lien entre la VOP et dp- ucMGP. Le modèle était ajusté pour l'âge, la fonction rénale et les risques CV classiques. Nous avons inclut 1001 participants dans les analyses (475 hommes et 526 femmes). La valeur moyenne de la VOP était de 7.87 ± 2.10 (m/s) et celle de dp-ucMGP de 0.43 ± 0.20 (nmol/L). La VOP était positivement et significativement associée à dp-ucMGP avant comme après ajustement pour le sexe, l'âge, l'indice de masse corporel, la taille, la pression artérielle systolique et diastolique, la fréquence cardiaque, la fonction rénale, les taux de cholestérol (LDL, HDL), la glycémie, la consommation de tabac, la présence d'un diabète, l'utilisation de médicaments antihypertenseurs ou hypolipémiants et la présence d'antécédents CV (P<0.01). En conclusion, des taux élevés de dp-ucMGP sont positivement et indépendamment associés à la rigidité artérielle après ajustement pour les facteurs de risques CV traditionnels, la fonction rénale et l'âge. Des études expérimentales sont nécessaires afin de déterminer si une supplémentation en vitamine K permet de ralentir l'avancement de la rigidité artérielle grâce à son activation de la MGP.

Investigating the functions of subregions within anterior hippocampus.

Previous functional MRI (fMRI) studies have associated anterior hippocampus with imagining and recalling scenes, imagining the future, recalling autobiographical memories and visual scene perception. We have observed that this typically involves the medial rather than the lateral portion of the anterior hippocampus. Here, we investigated which specific structures of the hippocampus underpin this observation. We had participants imagine novel scenes during fMRI scanning, as well as recall previously learned scenes from two different time periods (one week and 30 min prior to scanning), with analogous single object conditions as baselines. Using an extended segmentation protocol focussing on anterior hippocampus, we first investigated which substructures of the hippocampus respond to scenes, and found both imagination and recall of scenes to be associated with activity in presubiculum/parasubiculum, a region associated with spatial representation in rodents. Next, we compared imagining novel scenes to recall from one week or 30 min before scanning. We expected a strong response to imagining novel scenes and 1-week recall, as both involve constructing scene representations from elements stored across cortex. By contrast, we expected a weaker response to 30-min recall, as representations of these scenes had already been constructed but not yet consolidated. Both imagination and 1-week recall of scenes engaged anterior hippocampal structures (anterior subiculum and uncus respectively), indicating possible roles in scene construction. By contrast, 30-min recall of scenes elicited significantly less activation of anterior hippocampus but did engage posterior CA3. Together, these results elucidate the functions of different parts of the anterior hippocampus, a key brain area about which little is definitely known.

Systematic evaluation of matrix effects in hydrophilic interaction chromatography versus reversed phase liquid chromatography coupled to mass spectrometry.

Reversed phase liquid chromatography (RPLC) coupled to mass spectrometry (MS) is the gold standard technique in bioanalysis. However, hydrophilic interaction chromatography (HILIC) could represent a viable alternative to RPLC for the analysis of polar and/or ionizable compounds, as it often provides higher MS sensitivity and alternative selectivity. Nevertheless, this technique can be also prone to matrix effects (ME). ME are one of the major issues in quantitative LC-MS bioanalysis. To ensure acceptable method performance (i.e., trueness and precision), a careful evaluation and minimization of ME is required. In the present study, the incidence of ME in HILIC-MS/MS and RPLC-MS/MS was compared for plasma and urine samples using two representative sets of 38 pharmaceutical compounds and 40 doping agents, respectively. The optimal generic chromatographic conditions in terms of selectivity with respect to interfering compounds were established in both chromatographic modes by testing three different stationary phases in each mode with different mobile phase pH. A second step involved the assessment of ME in RPLC and HILIC under the best generic conditions, using the post-extraction addition method. Biological samples were prepared using two different sample pre-treatments, i.e., a non-selective sample clean-up procedure (protein precipitation and simple dilution for plasma and urine samples, respectively) and a selective sample preparation, i.e., solid phase extraction for both matrices. The non-selective pretreatments led to significantly less ME in RPLC vs. HILIC conditions regardless of the matrix. On the contrary, HILIC appeared as a valuable alternative to RPLC for plasma and urine samples treated by a selective sample preparation. Indeed, in the case of selective sample preparation, the compounds influenced by ME were different in HILIC and RPLC, and lower and similar ME occurrence was generally observed in RPLC vs. HILIC for urine and plasma samples, respectively. The complementary of both chromatographic modes was also demonstrated, as ME was observed only scarcely for urine and plasma samples when selecting the most appropriate chromatographic mode.

The ESRg matrix for strong field d5 systems

This review has tried to collect and correlate all the various equations for the g matrix of strong field d5 systems obtained from different basis sets using full electron and hole formalism calculations. It has corrected mistakes found in the literature and shown how the failure to properly take in symmetry boundary conditions has produced a variety of apparently inconsistent equations in the literature. The review has reexamined the problem of spin-orbit interaction with excited t4e states and finds that the earlier reports that it is zero in octahedral symmetry is not correct. It has shown how redefining what x, y, and z are in the principal coordinate system simplifies, compared to previous methods, the analysis of experimental g values with the equations.

Estimating extreme value cumulative distribution functions using bias-corrected kernel approaches

We propose a new kernel estimation of the cumulative distribution function based on transformation and on bias reducing techniques. We derive the optimal bandwidth that minimises the asymptotic integrated mean squared error. The simulation results show that our proposed kernel estimation improves alternative approaches when the variable has an extreme value distribution with heavy tail and the sample size is small.

On the Connectivity of the Julia sets of meromorphic functions

We prove that every transcendental meromorphic map $f$ with disconnected Julia set has a weakly repelling fixed point. This implies that the Julia set of Newton's method for finding zeroes of an entire map is connected. Moreover, extending a result of Cowen for holomorphic self-maps of the disc, we show the existence of absorbing domains for holomorphic self-maps of hyperbolic regions, whose iterates tend to a boundary point. In particular, the results imply that periodic Baker domains of Newton's method for entire maps are simply connected, which solves a well-known open question.