Bayesian nonparametric latent variable models

L’un des problèmes importants en apprentissage automatique est de déterminer la complexité du modèle à apprendre. Une trop grande complexité mène au surapprentissage, ce qui correspond à trouver des structures qui n’existent pas réellement dans les données, tandis qu’une trop faible complexité mène au sous-apprentissage, c’est-à-dire que l’expressivité du modèle est insuffisante pour capturer l’ensemble des structures présentes dans les données. Pour certains modèles probabilistes, la complexité du modèle se traduit par l’introduction d’une ou plusieurs variables cachées dont le rôle est d’expliquer le processus génératif des données. Il existe diverses approches permettant d’identifier le nombre approprié de variables cachées d’un modèle. Cette thèse s’intéresse aux méthodes Bayésiennes nonparamétriques permettant de déterminer le nombre de variables cachées à utiliser ainsi que leur dimensionnalité. La popularisation des statistiques Bayésiennes nonparamétriques au sein de la communauté de l’apprentissage automatique est assez récente. Leur principal attrait vient du fait qu’elles offrent des modèles hautement flexibles et dont la complexité s’ajuste proportionnellement à la quantité de données disponibles. Au cours des dernières années, la recherche sur les méthodes d’apprentissage Bayésiennes nonparamétriques a porté sur trois aspects principaux : la construction de nouveaux modèles, le développement d’algorithmes d’inférence et les applications. Cette thèse présente nos contributions à ces trois sujets de recherches dans le contexte d’apprentissage de modèles à variables cachées. Dans un premier temps, nous introduisons le Pitman-Yor process mixture of Gaussians, un modèle permettant l’apprentissage de mélanges infinis de Gaussiennes. Nous présentons aussi un algorithme d’inférence permettant de découvrir les composantes cachées du modèle que nous évaluons sur deux applications concrètes de robotique. Nos résultats démontrent que l’approche proposée surpasse en performance et en flexibilité les approches classiques d’apprentissage. Dans un deuxième temps, nous proposons l’extended cascading Indian buffet process, un modèle servant de distribution de probabilité a priori sur l’espace des graphes dirigés acycliques. Dans le contexte de réseaux Bayésien, ce prior permet d’identifier à la fois la présence de variables cachées et la structure du réseau parmi celles-ci. Un algorithme d’inférence Monte Carlo par chaîne de Markov est utilisé pour l’évaluation sur des problèmes d’identification de structures et d’estimation de densités. Dans un dernier temps, nous proposons le Indian chefs process, un modèle plus général que l’extended cascading Indian buffet process servant à l’apprentissage de graphes et d’ordres. L’avantage du nouveau modèle est qu’il admet les connections entres les variables observables et qu’il prend en compte l’ordre des variables. Nous présentons un algorithme d’inférence Monte Carlo par chaîne de Markov avec saut réversible permettant l’apprentissage conjoint de graphes et d’ordres. L’évaluation est faite sur des problèmes d’estimations de densité et de test d’indépendance. Ce modèle est le premier modèle Bayésien nonparamétrique permettant d’apprendre des réseaux Bayésiens disposant d’une structure complètement arbitraire.

Learning Binary Code Representations for Effective and Efficient Image Retrieval

The size of online image datasets is constantly increasing. Considering an image dataset with millions of images, image retrieval becomes a seemingly intractable problem for exhaustive similarity search algorithms. Hashing methods, which encodes high-dimensional descriptors into compact binary strings, have become very popular because of their high efficiency in search and storage capacity. In the first part, we propose a multimodal retrieval method based on latent feature models. The procedure consists of a nonparametric Bayesian framework for learning underlying semantically meaningful abstract features in a multimodal dataset, a probabilistic retrieval model that allows cross-modal queries and an extension model for relevance feedback. In the second part, we focus on supervised hashing with kernels. We describe a flexible hashing procedure that treats binary codes and pairwise semantic similarity as latent and observed variables, respectively, in a probabilistic model based on Gaussian processes for binary classification. We present a scalable inference algorithm with the sparse pseudo-input Gaussian process (SPGP) model and distributed computing. In the last part, we define an incremental hashing strategy for dynamic databases where new images are added to the databases frequently. The method is based on a two-stage classification framework using binary and multi-class SVMs. The proposed method also enforces balance in binary codes by an imbalance penalty to obtain higher quality binary codes. We learn hash functions by an efficient algorithm where the NP-hard problem of finding optimal binary codes is solved via cyclic coordinate descent and SVMs are trained in a parallelized incremental manner. For modifications like adding images from an unseen class, we propose an incremental procedure for effective and efficient updates to the previous hash functions. Experiments on three large-scale image datasets demonstrate that the incremental strategy is capable of efficiently updating hash functions to the same retrieval performance as hashing from scratch.

"Right to Personal Liberty Under the Indian Constitution - With Special Reference to Judicial Process"

The nature and extent of protection secured to personal liberty has been a subject matter of great controversy and debate. The expression "procedure established by law" as a standard of protection for personal liberty has been looked upon as highly unsatisfactory and inadequate. For, unlike the specific attributes of liberty that are separately guaranteed under Art.19, ‘personal liberty‘ as guaranteed by Art.21 does not obligate the .Legislature to comply with the requirements of justice and reasonableness as and when it enchroaches upon that right. Though the concept of ‘personal liberty‘ has received an evolutive and expansive meaning through judicial process, the standard of protection which the judicial process could secure to personal liberty through the interpretation of Art.21 has been far from satisfactory Even after four decades of judicial process in the interpretation of Art.21 the problem of evolving a just and adequate standard of protection for personal liberty in that Article continues to be 21 crucial constitutional issue, craving for a. satisfactory solution. And the present study is a humble attempt to unravel this problem and to Search for a reasonable solution.

Arbitrage Pricing Theory and Return Generating Process: A Macroeconomic Approach to Indian Stock Market

A study focusing on the identification of return generating factors and to the extent of their influence on share prices the outcome will be a tool for investment analysis in the hands of investors portfolio managers and mutual funds who are mostly concerned with changing share prices. Since the study takes into account the influence of macroeconomic variables on variations in share returns by using the outcome the government can frame out suitable policies on long term basis and that will help in nurturing a healthy economy and resultant stock market. As every company management tries to maximize the wealth of the share holders a clear idea about the return generating variables and their influence will help the management to frame various policies to maximize the wealth of the shareholders.

Is the Indian Ocean SST variability a homogeneous diffusion process.

Whereas the predominance of El Niño Southern Oscillation (ENSO) mode in the tropical Pacific sea surface temperature (SST) variability is well established, no such consensus seems to have been reached by climate scientists regarding the Indian Ocean. While a number of researchers think that the Indian Ocean SST variability is dominated by an active dipolar-type mode of variability, similar to ENSO, others suggest that the variability is mostly passive and behaves like an autocorrelated noise. For example, it is suggested recently that the Indian Ocean SST variability is consistent with the null hypothesis of a homogeneous diffusion process. However, the existence of the basin-wide warming trend represents a deviation from a homogeneous diffusion process, which needs to be considered. An efficient way of detrending, based on differencing, is introduced and applied to the Hadley Centre ice and SST. The filtered SST anomalies over the basin (23.5N-29.5S, 30.5E-119.5E) are then analysed and found to be inconsistent with the null hypothesis on intraseasonal and interannual timescales. The same differencing method is then applied to the smaller tropical Indian Ocean domain. This smaller domain is also inconsistent with the null hypothesis on intraseasonal and interannual timescales. In particular, it is found that the leading mode of variability yields the Indian Ocean dipole, and departs significantly from the null hypothesis only in the autumn season.

Knowledge spillover in Indian automobile industry : the process and the coverage

In India, as the production of passenger cars increased, many local small and medium enterprises (SMEs) entered the parts and components manufacturing sector. The sources of knowledge for large enterprises and SMEs are different. Naturally, spillover effects among large enterprises and between large enterprises and SMEs are different. This paper focuses on knowledge spillover among large enterprises and from large enterprises to SMEs. Subcontractor can absorb relation-specific skills through repeated interaction with parent company. The results of field survey emphasizes that relation-specific skills are a determinant factor of spillover effects from assemblers and large auto component manufacturers to SMEs. Econometric analysis shows that spillover effects among medium and large automobile units and from medium and large automobile units to small units went beyond boundary of cluster.

The camera; or, Art of drawing in water colours : with instructions for sketching form nature : comprising the whole process of water-coloured drawing, familiarly exmplified in drawing, shadowing, and tinting a complete landscape ... and directions for compounding and using colours, sepia, Indian ink, bister, &c.

Mode of access: Internet.

Analytic hierarchy process helps evaluate projects in Indian oil pipeline industry

Conventionally, oil pipeline projects are evaluated thoroughly by the owner before investment decision is made using market, technical and financial analysis sequentially. The market analysis determines pipelines throughput and supply and demand points. Subsequent, technical analysis identifies technological options and economic and financial analysis then derives the least cost option among all technically feasible options. The subsequent impact assessment tries to justify the selected option by addressing environmental and social issues. The impact assessment often suggests alternative sites, technologies, and/or implementation methodology, necessitating revision of technical and financial analysis. This study addresses these issues via an integrated project evaluation and selection model. The model uses analytic hierarchy process, a multiple-attribute decision-making technique. The effectiveness of the model has been demonstrated through a case application on cross-country petroleum pipeline project in India.

International partner selection Indian technical textile market

Due to a combination of a vast agricultural industry and a tremendously growing technical textile industry, Ludvig Svensson identified India as target market for possible expansion through domestic production and supply. However, Svensson needed additional information about the industry structure and key players. Therefore, this project focused on a detailed analysis of the technical textile market and its players by following the international partner selection process. Thereby, five key players were identified as potential partners, as well as the need for additional research to determine alternative entry modes, as the market does not currently seem to be receptive for Svensson products.

Strategic alliances as an international entry strategy: Finnish cleantech SMEs and the Indian market

The demand for environmental technologies, also called cleantech, is growing globally but the need is especially high in emerging markets such as India where the rising economy and rapid industrialisation have led to increasing energy needs and environmental degradation. The market is of great potential also for the Finnish cleantech cluster that represents advanced expertise in several fields of environmental technologies. However, most of the Finnish companies in the field are SMEs that face challenges in their internationalisation due to their limited resources. The objective of this study was to estimate, whether strategic alliances could be an efficient entry strategy for Finnish cleantech SMEs entering the Indian market. This was done by studying what are the key factors influencing the international entry mode decision of Finnish cleantech SMEs, what are the major factors affecting the entry of Finnish cleantech SMEs to the Indian market and how do Finnish cleantech SMEs use strategic alliances in their internationalisation process. The study was realised as a qualitative multi-case study through theme interviews of Finnish cleantech SME representatives. The results indicated that Finnish cleantech SMEs prefer to enter international markets through non-equity and collaborative modes of entry. These entry modes are chosen because of the small size and limited resources of companies, but also because they want to protect their innovative technologies from property rights violations. India is an attracting market for Finnish cleantech SMEs mainly because of its size and growth, but insufficient environmental regulation and high import tariffs have hindered entry to the market. Finnish cleantech SMEs commonly use strategic alliances in their internationalisation process but the use is rather one-sided. Most of the formed strategic alliances are low-commitment, international contractual agreement in sales and distribution. Alliance partner selection receives less attention. In the future, providing Finnish cleantech SMEs with international experience and training could help in diversifying the use of strategic alliances and increase their benefits to SME internationalisation.

Finnish buyers’ perceptions of Indian textile and clothing suppliers: Country of origin in focus

Choosing the right supplier is crucial for long-term business prospects and profitability. Thus organizational buyers are naturally very interested in how they can select the right supplier for their needs. Likewise, suppliers are interested in knowing how their customers make purchasing decisions in order to effectively sell and market to them. From the point of view of the textile and clothing (T&C) industry, regulatory changes and increasing low-cost and globalization pressures have led to the rise of low-cost production locations India and China as the world’s largest T&C producers. This thesis will examine T&C trade between Finland and India specifically in the context of non-industrial T&C products. Its main research problem asks: what perceptions do Finnish T&C industry buyers hold of India and Indian suppliers? B2B buyers use various supplier selection models and criteria in making their purchase decisions. A significant amount of research has been done into supplier selection practices, and in the context of international trade, country of origin (COO) perceptions specifically have garnered much attention. This thesis uses a mixed methods approach (online questionnaire and in-depth interviews) to evaluate Finnish T&C buyers’ supplier selection criteria, COO perceptions of India and experiences of Indian suppliers. It was found that the most important supplier selection criteria used by Finnish T&C buyers are quality, reliability and cost. COO perceptions were not found to be influential in purchasing process. Indian T&C suppliers’ strengths were found to be low cost, flexibility and a history of traditional T&C expertise. Their weaknesses include product quality and unreliable delivery times. Overall, the main challenges that need to be overcome by Indian T&C companies are logistical difficulties and the cost vs. quality trade-off. Despite positive perceptions of India for cost, the overall value offered by Indian T&C products was perceived to be low due to poor quality. Unreliable delivery time experiences also affected buyer’s reliability perceptions of Indian suppliers. The main limiting factors of this thesis relate to the small sample size used in the research. This limits the generalizability of results and the ability to evaluate the reliability and validity of some of the research instruments.

Only if “je est un autre” Can I Recognise You: Reflections on Canada’s Process of Constitutional Recognition of the “Preexistence of Distinctive Aborignial Societies”

1er Prix du concours d'initiation à la recherche organisé par le Regroupement Droit et Changements. La Loi sur les Indiens institutionnalise toujours de nombreuses facettes de ce qu’est être « Indien » pour beaucoup d’individus au Canada et un changement de perspective doit être opéré. Cet essai puise dans la pensée du philosophe Theodor Adorno pour réfléchir aux tentatives de reconnaissance juridique par le Canada des individus et sociétés autochtones en vertu de l’article 35 de la Constitution. L’auteur présente la théorie de la dialectique négative d’Adorno de 1966 sur le rapport à l’altérité, à partir de l’analyse de la professeure Drucilla Cornell, afin d’identifier ce que sa pensée prescrit pour établir des rapports non-oppressants entre Autochtones et non-Autochtones et leurs gouvernements aujourd’hui. La dialectique négative est particulièrement appropriée à la tentative de reconnaissance juridique de l’existence des sociétés autochtones par le Canada, du fait de leur statut marginalisé et de leurs revendications à la spécificité. Après avoir établi un tel cadre, l’auteur souligne que des précédentes tentatives de reconnaissances se sont soldées par des échecs en raison des désaccords au niveau des valeurs impliquées et des concepts utilisés auxquels elles ont donné lieu. Le processus de signature des traités numérotés de 1871-1921 est employé comme illustration en raison de son résultat souvent décrit aujourd’hui comme coercitif et injuste en dépit du discours de négociation sur un pied d’égalité l’ayant accompagné. Les critiques contemporaines de la politique en vigueur de mise en œuvre de l’autonomie gouvernementale autochtone par des accords négociés sont également présentées, afin d’illustrer que des désaccords quant à la manière dont l’État canadien entend reconnaître les peuples autochtones persistent à ce jour. L’auteur ajoute que, du point de vue de la dialectique négative, de tels désaccords doivent nécessairement être résolus pour que des rapports moins oppressifs puissent être établis. L’auteur conclut que la dialectique négative impose à la fois de se considérer soi-même (« je est un autre ») et de considérer l’autre comme au-delà des limites de sa propre pensée. La Cour suprême a déjà reconnu que la seule perspective de la common law n’est pas suffisante pour parvenir à une réconciliation des souverainetés des Autochtones et de la Couronne en vertu de la Constitution. Le concept de common law de fiduciaire présente un véhicule juridique intéressant pour une reconfiguration plus profonde par le gouvernement canadien de son rapport avec les peuples autochtones, priorisant processus plutôt que résultats et relations plutôt que certitude. Il doit toutefois être gardé à l’esprit que la reconnaissance de ces peuples par l’État canadien par le prisme de la pensée d’Adorno présente non seulement le défi d’inclure de nouvelles perspectives, mais également de remettre en cause les prémisses fondamentales à partir desquelles on considère la communauté canadienne en général.

Sea Surface Temperature-Convection Relationship in Tropical Oceans with Special Emphasis to Intraseasonal Variability of Indian Summer Monsoon

The SST convection relation over tropical ocean and its impact on the South Asian monsoon is the first part of this thesis. Understanding the complicated relation between SST and convection is important for better prediction of the variability of the Indian monsoon in subseasonal, seasonal, interannual, and longer time scales. Improved global data sets from satellite scatterometer observations of SST, precipitation and refined reanalysis of global wind fields have made it possible to do a comprehensive study of the SST convection relation. Interaction of the monsoon and Indian ocean has been discussed. A coupled feedback process between SST and the Active-Break cycle of the Asian summer monsoon is a central theme of the thesis. The relation between SST and convection is very important in the field of numerical modeling of tropical rainfall. It is well known that models generally do very well simulating rainfall in areas of tropical convergence zones but are found unable to do satisfactory simulation in the monsoon areas. Thus in this study we critically examined the different mechanisms of generation of deep convection over these two distinct regions.The study reported in chapter 3 has shown that SST - convection relation over the warm pool regions of Indian and west Pacific oceans (monsoon areas) is in such a way that convection increases with SST in the SST range 26-29 C and for SST higher than 29-30 C convection decreases with increase of SST (it is called Waliser type). It is found that convection is induced in areas with SST gradients in the warm pool areas of Indian and west Pacific oceans. Once deep convection is initiated in the south of the warmest region of warm pool, the deep tropospheric heating by the latent heat released in the convective clouds produces strong low level wind fields (Low level Jet - LLJ) on the equatorward side of the warm pool and both the convection and wind are found to grow through a positive feedback process. Thus SST through its gradient acts only as an initiator of convection. The central region of the warm pool has very small SST gradients and large values of convection are associated with the cyclonic vorticity of the LLJ in the atmospheric boundary layer. The conditionally unstable atmosphere in the tropics is favorable for the production of deep convective clouds.

A Study on the Transport and Distribution of Atmospheric Aerosols and its Influence on Regional Meteorological Parameters over the Indian Subcontinent

This doctoral thesis addresses the growing concern about the significant changes in the climatic and weather patterns due to the aerosol loading that have taken place in the Indo Gangetic Plain(IGP)which includes most of the Northern Indian region. The study region comprises of major industrial cities in India (New Delhi, Kanpur, Allahabad, Jamshedpur and Kolkata). Northern and central parts of India are one of the most thickly populated areas in the world and have the most intensely farmed areas. Rapid increase in population and urbanization has resulted in an abrupt increase in aerosol concentrations in recent years. The IGP has a major source of coal; therefore most of the industries including numerous thermal power plants that run on coal are located around this region. They inject copious amount of aerosols into the atmosphere. Moreover, the transport of dust aerosols from arid locations is prevalent during the dry months which increase the aerosol loading in theatmosphere. The topography of the place is also ideal for the congregation of aerosols. It is bounded by the Himalayas in the north, Thar Desert in the west, the Vindhyan range in the south and Brahmaputra ridge in the east. During the non‐monsoon months (October to May) the weather in the location is dry with very little rainfall. Surface winds are weak during most of the time in this dry season. The aerosols that reach the location by means of long distance transport and from regional sources get accumulated under these favourable conditions. The increase in aerosol concentration due to the complex combination of aerosol transport and anthropogenic factors mixed with the contribution from the natural sources alters the optical properties and the life time of clouds in the region. The associated perturbations in radiative balance have a significant impact on the meteorological parameters and this in turn determines the precipitation forming process. Therefore, any change in weather which disturbs the normal hydrological pattern is alarming in the socio‐economic point of view. Hence, the main focus of this work is to determine the variation in transport and distribution of aerosols in the region and to understand the interaction of these aerosols with meteorological parameters and cloud properties.