Modèles de Markov cachés à haute précision dynamique

La reconnaissance vocale est une technologie sujette à amélioration. Malgré 40 ans de travaux, de nombreuses applications restent néanmoins hors de portée en raison d'une trop faible efficacité. De façon à pallier à ce problème, l'auteur propose une amélioration au cadre conceptuel classique. Plus précisément, une nouvelle méthode d'entraînement des modèles markoviens cachés est exposée de manière à augmenter la précision dynamique des classificateurs. Le présent document décrit en détail le résultat de trois ans de recherche et les contributions scientifiques qui en sont le produit. L'aboutissement final de cet effort est la production d'un article de journal proposant une nouvelle tentative d'approche à la communauté scientifique internationale. Dans cet article, les auteurs proposent que des topologies finement adaptées de modèles markoviens cachés (HMMs) soient essentielles à une modélisation temporelle de haute précision. Un cadre conceptuel pour l'apprentissage efficace de topologies par élagage de modèles génériques complexes est donc soumis. Des modèles HMM à topologie gauche-à-droite sont d'abord entraînés de façon classique. Des modèles complexes à topologie générique sont ensuite obtenus par écrasement des modèles gauche-à-droite. Finalement, un enchaînement successif d'élagages et d'entraînements Baum-Welch est fait de manière à augmenter la précision temporelle des modèles.

Markov Random Fields in Cancer Mutation Dependencies

Even though a large amount of evidence would suggest that PP2A serine/threonine protein phosphatase acts as a tumour suppressor the genomics data to support this claim is limited. We fit a sparse binary Markov random field with individual sample's total mutational frequency as an additional covariate to model the dependencies between the mutations occurring in the PP2A encoding genes. We utilize the data from recent large scale cancer genomics studies, where the whole genome from a human tumour biopsy has been analysed. Our results show a complex network of interactions between the occurrence of mutations in our twenty examined genes. According to our analysis the mutations occurring in the genes PPP2R1A, PPP2R3A, and PPP2R2B are identified as the key mutations. These genes form the core of the network of conditional dependency between the mutations in the investigated twenty genes. Additionally, we note that the mutations occurring in PPP2R4 seem to be more influential in samples with higher number of total mutations. The mutations occurring in the set of genes suggested by our results has been shown to contribute to the transformation of human cells. We conclude that our evidence further supports the claim that PP2A acts as a tumour suppressor and restoring PP2A activity is an appealing therapeutic strategy.

Structural characterization of interaction of K11 and K63 mixed-linkage polyubiquitin chains with the tUIMs of epsin1

Ubiquitylation or covalent attachment of ubiquitin (Ub) to a variety of substrate proteins in cells is a versatile post-translational modification involved in the regulation of numerous cellular processes. The distinct messages that polyubiquitylation encodes are attributed to the multitude of conformations possible through attachment of ubiquitin monomers within a polyubiquitin chain via a specific lysine residue. Thus the hypothesis is that linkage defines polyubiquitin conformation which in turn determines specific recognition by cellular receptors. Ubiquitylation of membrane surface receptor proteins plays a very important role in regulating receptor-mediated endocytosis as well as endosomal sorting for lysosomal degradation. Epsin1 is an endocytic adaptor protein with three tandem UIMs (Ubiquitin Interacting Motifs) which are responsible for the highly specific interaction between epsin and ubiquitylated receptors. Epsin1 is also an oncogenic protein and its expression is upregulated in some types of cancer. Recently it has been shown that novel K11 and K63 mixed-linkage polyubiquitin chains serve as internalization signal for MHC I (Major Histocompatibility Complex I) molecule through their association with the tUIMs of epsin1. However the molecular mode of action and structural details of the interaction between polyubiquitin chains on receptors and tUIMs of epsin1 is yet to be determined. This information is crucial for the development of anticancer therapeutics targeting epsin1. The molecular basis for the linkage-specific recognition of K11 and K63 mixed-linkage polyubiquitin chains by the tandem UIMs of the endocytic adaptor protein epsin1 is investigated using a combination of NMR methods.

Two novel COLVI long chains in zebrafish that are essential for muscle development

Collagen VI (COLVI), a protein ubiquitously expressed in connective tissues, is crucial for structural integrity, cellular adhesion, migration and survival. Six different genes are recognized in mammalians, encoding six COLVI-chains that assemble as two ‘short’ (α1, α2) and one ‘long’ chain (theoretically any one of α3–6). In humans, defects in the most widely expressed heterotrimer (α123), due to mutations in the COL6A1-3 genes, cause a heterogeneous group of neuromuscular disorders, collectively termed COLVI-related muscle disorders. Little is known about the function(s) of the recently described α4-6 chains and no mutations have been detected yet. In this study, we characterized two novel COLVI long chains in zebrafish that are most homologous to the mammalian α4 chain; therefore, we named the corresponding genes col6a4a and col6a4b. These orthologues represent ancestors of the mammalian Col6a4-6 genes. By in situ hybridization and RT-qPCR, we unveiled a distinctive expression kinetics for col6a4b, compared with the other col6a genes. Using morpholino antisense oligonucleotides targeting col6a4a, col6a4b and col6a2, we modelled partial and complete COLVI deficiency, respectively. All morphant embryos presented altered muscle structure and impaired motility. While apoptosis was not drastically increased, autophagy induction was defective in all morphants. Furthermore, motoneuron axon growth was abnormal in these morphants. Importantly, some phenotypical differences emerged between col6a4a and col6a4b morphants, suggesting only partial functional redundancy. Overall, our results further confirm the importance of COLVI in zebrafish muscle development and may provide important clues for potential human phenotypes associated with deficiency of the recently described COLVI-chains.

Management accounting and supply chains: actions, concerns, and networks

What does this thesis do? This thesis uses Actor-Network Theory (ANT) to examine how a UK retailer’s organization and strategy, and, in turn, its form of management accounting was shaped by its supply chain. The thesis does this by reporting on four related themes in the form of four inter-connected essays. The first essay undertakes a state-of-the-art review of the literature. It examines how accounting issues within supply chains permeate ‘matters of concern’. In accordance with this idea of ANT, the essay illustrates how issues emerged, controversies developed, and matters evolved through an actor-network of accounting researchers within the supply chain domain. This leads on to the second essay, which exemplifies the nature of the UK’s retailing industry within which the supply chain case organization emerged and developed. The purposes of the essay are twofold: to introduce the contextual ramifications of the case organization; and to illustrate the emergence of a new market logic, which led to the creation of a global supply chain and a new form of management accounting therein. The third essay reports on a qualitative case study. It analyses the dualistic relation between ostensive and performative aspects of supply chain strategy, reveals how accounting numbers act as an obligatory passage point within this dualism, and makes a contribution to the ANT debate around the issue of whether and how a dualism between ostensive and performative aspects exists. The final essay reports on another case analysis of institutionalizing a heterarchical form of management accounting: a distributed form of intelligence that penetrates through lateral accountable relations. The analysis reveals a new form of management accounting characterised by ambiguity; it emphasizes the possibilities of compromises and negotiations, and it thus contributes to knowledge by combining an aspect of ANT with heterarchical tendencies in the world of contemporary organizations. Finally, the thesis concludes that it is the supply chain that organises today’s neoliberal capitalism; and it is management accounting that unites both human and non-human actors within such supply chains, despite that form of management accounting being ambiguous. The thesis comprises the introduction, these four essays, and the conclusion.

Value chain transformation: Taking stock of WorldFish research on value chains and markets

The goal of WorldFish’s research on markets and value chains is to increase the benefits to resource-poor people from fisheries and aquaculture value chains by researching (1) key barriers to resource-poor men, women and other marginalized groups gaining greater benefits from participation in value chains, including barriers related to the availability, affordability and quality of nutrient-rich fish for resource-poor consumers; (2) interventions to overcome those barriers; and (3) mechanisms that are most effective for scaling up of value chain interventions. This paper aims to promote and document learning across WorldFish’s value chain research efforts in Asia and Africa. It has three main objectives: (1) to take stock of WorldFish’s past and ongoing research on value chains; (2) to draw out commonalities and differences between these projects; and (3) to provide a synthesis of some learning that can guide future work.

CHARACTERIZATION OF POLYUBIQUITIN CHAINS BY MASS SPECTROMETRY

The work outlined in this dissertation will allow biochemists and cellular biologists to characterize polyubiquitin chains involved in their cellular environment by following a facile mass spectrometric based workflow. The characterization of polyubiquitin chains has been of interest since their discovery in 1984. The profound effects of ubiquitination on the movement and processing of cellular proteins depend exclusively on the structures of mono and polyubiquitin modifications anchored or unanchored on the protein within the cellular environment. However, structure-function studies have been hindered by the difficulty in identifying complex chain structures due to limited instrument capabilities of the past. Genetic mutations or reiterative immunoprecipitations have been used previously to characterize the polyubiquitin chains, but their tedium makes it difficult to study a broad ubiquitinome. Top-down and middle-out mass spectral based proteomic studies have been reported for polyubiquitin and have had success in characterizing parts of the chain, but no method to date has been successful at differentiating all theoretical ubiquitin chain isomers (ubiquitin chain lengths from dimer to tetramer alone have 1340 possible isomers). The workflow presented here can identify chain length, topology and linkages present using a chromatographic-time-scale compatible, LC-MS/MS based workflow. To accomplish this feat, the strategy had to exploit the most recent advances in top-down mass spectrometry. This included the most advanced electron transfer dissociation (ETD) activation and sensitivity for large masses from the orbitrap Fusion Lumos. The spectral interpretation had to be done manually with the aid of a graphical interface to assign mass shifts because of a lack of software capable to interpret fragmentation across isopeptide linkages. However, the method outlined can be applied to any mass spectral based system granted it results in extensive fragmentation across the polyubiquitin chain; making this method adaptable to future advances in the field.

A risk model with an observer in a Markov environment

We consider a spectrally-negative Markov additive process as a model of a risk process in a random environment. Following recent interest in alternative ruin concepts, we assume that ruin occurs when an independent Poissonian observer sees the process as negative, where the observation rate may depend on the state of the environment. Using an approximation argument and spectral theory, we establish an explicit formula for the resulting survival probabilities in this general setting. We also discuss an efficient evaluation of the involved quantities and provide a numerical illustration.

Service Orientation in Demand-Supply Chains: Towards an Integrated Framework

Part 5: Service Orientation in Collaborative Networks

Synchronization of Heteroclinic Circuits Through Learning in Chains of Neural Motifs

The synchronization of oscillatory activity in networks of neural networks is usually implemented through coupling the state variables describing neuronal dynamics. In this study we discuss another but complementary mechanism based on a learning process with memory. A driver network motif, acting as a teacher, exhibits winner-less competition (WLC) dynamics, while a driven motif, a learner, tunes its internal couplings according to the oscillations observed in the teacher. We show that under appropriate training the learner motif can dynamically copy the coupling pattern of the teacher and thus synchronize oscillations with the teacher. Then, we demonstrate that the replication of the WLC dynamics occurs for intermediate memory lengths only. In a unidirectional chain of N motifs coupled through teacher-learner paradigm the time interval required for pattern replication grows linearly with the chain size, hence the learning process does not blow up and at the end we observe phase synchronized oscillations along the chain. We also show that in a learning chain closed into a ring the network motifs come to a consensus, i.e. to a state with the same connectivity pattern corresponding to the mean initial pattern averaged over all network motifs.

Foreign direct investment determinants revisited in the context of global value chains

In this paper, we aim at contributing to the new field of research that intends to bring up-to-date the tools and statistics currently used to look to the current reality given by Global Value Chains (GVC) in international trade and Foreign Direct Investment (FDI). Namely, we make use of the most recent data published by the World Input-Output Database to suggest indicators to measure the participation and net gains of countries by being a part of GVC; and use those indicators in a pooled-regression model to estimate determinants of FDI stocks in Organization for Economic Co-operation and Development (OECD)-member countries. We conclude that one of the measures proposed proves to be statistically significant in explaining the bilateral stock of FDI in OECD countries, meaning that the higher the transnational income generated between two given countries by GVC, taken as a proxy to the participation of those countries in GVC, the higher one could expect the FDI entering those countries to be. The regression also shows the negative impact of the global financial crisis that started in 2009 in the world’s bilateral FDI stocks and, additionally, the particular and significant role played by the People’s Republic of China in determining these stocks.

Mapping agricultural value chains with international input-output data

In recent years, the analysis of trade in value added has been explored by many researchers. Although they have made important contributions by developing GVC-related indices and proposing techniques for decomposing trade data, they have not yet explored the method of value chain mapping—a core element of conventional value chain analysis. This paper introduces a method of value chain mapping that uses international input-output data and reveals both upstream and downstream transactions of goods and services induced by production activities of a specific commodity or industry. This method is subsequently applied to the agricultural value chain of three Greater Mekong Sub-region countries (i.e., Thailand, Vietnam, and Cambodia). The results show that the agricultural value chain has been increasingly internationalized, although there is still room for obtaining benefits from GVC participation, especially in a country such as Cambodia.

HOG CHAINS AND MARK TWAINS: A STUDY OF LABOR HISTORY, ARCHAEOLOGY, AND INDUSTRIAL ETHNOGRAPHY OF THE STEAMBOAT ERA OF THE MONONGAHELA VALLEY 1811-1950

This dissertation examines a unique working class in the United States, the men and women who worked on the steamboats from the Industrial Revolution until the demise of steam-powered boats in the mid-20th century. The steamboat was the beginning of a technological system that was developed in America and used in such great numbers that it made the rapid population of the Trans-Appalachian West possible. The steamboat was forever romanticized by images of the antebellum South or the quick wit of Samuel Clemens and his sentimental book, Life on the Mississippi. The imagination swirls with thoughts of boats, bleach white, slowly churning the calm waters of some Spanish moss covered river. The reality of the boats and the experience of those who worked on them has been lost in this nostalgic vision. This research details the history of the western steamboat in the Monongahela Valley, the birthplace of the commercial steamboat industry. The first part of this dissertation examines the literature of authors in the field of labor history and Industrial Archaeology to place this work into the larger context of published literature. The second builds a framework for understanding the various eras that the steamboat went through both in terms of technological change, but also the change the workers experienced as their identity as a working class was being shaped. The third part details the excavations of two steamboat captains houses, those of Captain James Gormley and Captain Michael A. Cox. Both men represented a time in which the steamboat was in an era of transition. Excavations at their homes yield clues to their class status and how integrated they were in the local community. The fourth part of this study documents the oral histories of steamboat workers, both men and women, and their experience on the boats and on the river. Their rapidly declining population of those who lived and worked on the boats gives urgency for their lives to be documented. Finally, this study concludes with a synthesis of how worker identity solidified in the face of technological, socio-economic, and ideological change especially during their push for unionization and the introduction of the diesel towboat.

A Markov switching unobserved component analysis of the CDX index term premium

Using a Markov switching unobserved component model we decompose the term premium of the North American CDX index into a permanent and a stationary component. We establish that the inversion of the CDX term premium is induced by sudden changes in the unobserved stationary component, which represents the evolution of the fundamentals underpinning the probability of default in the economy. We find evidence that the monetary policy response from the Fed during the crisis period was effective in reducing the volatility of the term premium. We also show that equity returns make a substantial contribution to the term premium over the entire sample period.

Participation of portuguese manufacturing industry in global value chains

Portugal’s manufacturing sector has a significant importance both in national income and employment. As has been pointed out by several researchers, the traditional methods of analysis fail to grasp all the dimensions of economic competitiveness. This dissertation is then, at its core, an analysis of Portugal’s manufacturing industry in terms of the latter’s value added to production and impact to employment under the framework of global value chains. The current dissertation seeks to study in which way the Portuguese manufacturing industry, and its respective sectors, has a direct and indirect impact on the creation of value added and employment and how this impact can be measured. For development of this work the input-output approach for calculation of multipliers and the new framework proposed by Timmer et al. (2013) for calculation of GVC income and GVC jobs indicators were used, elaborated on the basis of the WIOD project dataset. Moreover, to illustrate the application of the provided methodology the Portuguese textile industry was used as an example. It was found that the changes in final demand of such sectors as Pulp, Paper, Printing and Publishing; Machinery, Nec and Textiles and Textile Products would have a larger impact on generated value added than other manufacturing sectors. At the same time, employment created by the changes in final demand would be more impacted by such sectors as Food, Beverages and Tobacco; Wood and Products of Wood and Cork and Textiles and Textile Products. In this regard, the number of low-skilled workers in Portugal seems to be more effected by changes in final demand, than those occupied by higher -skilled individuals. Moreover, it was found that the distribution of GVC income and GVC jobs for the Portuguese manufacturing industry shares a similar outlook. However, upon closer inspection of GVC labour distribution by skill levels there seems to exist a general progression in which low-skilled jobs requirements are met by local resources, while the need for higher skilled jobs require a greater “off-shoring” of work The results obtained through calculations of presented multipliers provide a powerful tool for policy makers in strategic planning of development of national economy. Using the provided methodology and obtained results, a government and supranational organizations could define which industry would have the greatest impact for an additional unit of output generated through the economy, and thus define the sectors for further investments.