Leveraging noisy side information for disentangling of factors of variation in a supervised setting

Ce mémoire est composé de trois articles et présente les résultats de travaux de recherche effectués dans le but d'améliorer les techniques actuelles permettant d'utiliser des données associées à certaines tâches dans le but d'aider à l'entraînement de réseaux de neurones sur une tâche différente. Les deux premiers articles présentent de nouveaux ensembles de données créés pour permettre une meilleure évaluation de ce type de techniques d'apprentissage machine. Le premier article introduit une suite d'ensembles de données pour la tâche de reconnaissance automatique de chiffres écrits à la main. Ces ensembles de données ont été générés à partir d'un ensemble de données déjà existant, MNIST, auquel des nouveaux facteurs de variation ont été ajoutés. Le deuxième article introduit un ensemble de données pour la tâche de reconnaissance automatique d'expressions faciales. Cet ensemble de données est composé d'images de visages qui ont été collectées automatiquement à partir du Web et ensuite étiquetées. Le troisième et dernier article présente deux nouvelles approches, dans le contexte de l'apprentissage multi-tâches, pour tirer avantage de données pour une tâche donnée afin d'améliorer les performances d'un modèle sur une tâche différente. La première approche est une généralisation des neurones Maxout récemment proposées alors que la deuxième consiste en l'application dans un contexte supervisé d'une technique permettant d'inciter des neurones à apprendre des fonctions orthogonales, à l'origine proposée pour utilisation dans un contexte semi-supervisé.

A Study Of The Marketing Channels Of Natural Rubber With Special Reference To Co-Operative Marketing In Kerala

The study is designed to gather, record, analyse and critically evaluate data on natural rubber production, consumption and marketing in Kerala. The scope of the study also covers the processing of natural rubber by rubber growers, especially small growers. Taking into consideration Kerala’s supremacy in natural rubber cultivation, a detailed study of marketing channels of natural rubber in Kerala, by giving special reference to co-operative rubber marketing has been given importance. Attempt has also been made to collect secondary data of the last 15 years

Structure of marketing channels of selected manufacturing industries in Kerala

The study evolved from the basic premise that the existing distribution structure is not adequate or adaptive to meet the needs of the expanding manufacturing sector and the emerging mass market. The hypothised causes of the problem are the following: marketing channels are not used for strategy differentiation by manufacturers: there are too many intermediaries in the channels; the distributive institutions are not adaptive; and there is very little control over the flow of products through the channels. These assumptions about the causes of the problem have been translated into specific hypotheses and tested with data. Empirical analysis, while supporting some of these hypotheses, challenges certain widely held notions. The ensuing summary presents the important findings, in the sequence in which they are discussed in the study.

Marketing Channels in Ornamental Fish Trade in West Bengal

A lucrative export market and high domestic demand has made ornamental fish industry in West Bengal a potential source for income generation. The study aimed to identify: (i) the commercially important size groups of main ornamental fish varieties available in the state; (ii) the existing supply chain; (iii) major constraints for development of the industry; (iv) and to anlayse price spread of commercially important varieties; and (v) to evaluate the profitability of operation at different stakeholder levels in the marketing chain. Export market of ornamental fishes in the state followed a single supply channel while three different distribution channels existed in the domestic market. High electricity charges was the major problem faced by breeders (producers/rearers) whereas lack of technical knowledge regarding transportation was the major constraint for wholesalers. Lack of knowledge on proper health management inhibited the growth of retail industry. The fresh water catfish, angel, molly, arowana, gold fish, tetras, and gouramis showed comparatively higher breeders’ share in consumers’ rupee. Wholesalers were earning comparatively higher annual profit than the other stakeholders due to moderate initial investment and also due to the comparatively lower risk involved.

Fast Object Recognition in Noisy Images Using Simulated Annealing

A fast simulated annealing algorithm is developed for automatic object recognition. The normalized correlation coefficient is used as a measure of the match between a hypothesized object and an image. Templates are generated on-line during the search by transforming model images. Simulated annealing reduces the search time by orders of magnitude with respect to an exhaustive search. The algorithm is applied to the problem of how landmarks, for example, traffic signs, can be recognized by an autonomous vehicle or a navigating robot. The algorithm works well in noisy, real-world images of complicated scenes for model images with high information content.

"Functional Upregulation of Ca2+ -Activated K+ Channels in the Development of Substantia Nigra Dopamine Neurons"

Many connections in the basal ganglia are made around birth when animals are exposed to a host of new affective, cognitive, and sensori-motor stimuli. It is thought that dopamine modulates cortico-striatal synapses that result in the strengthening of those connections that lead to desired outcomes. We propose that there must be a time before which stimuli cannot be processed into functional connections, otherwise it would imply an effective link between stimulus, response, and reward in uterus. Consistent with these ideas, we present evidence that early in development dopamine neurons are electrically immature and do not produce high-frequency firing in response to salient stimuli. We ask first, what makes dopamine neurons immature? and second, what are the implications of this immaturity for the basal ganglia? As an answer to the first question, we find that at birth the outward current is small (3nS-V), insensitive to Ca2z, TEA, BK, and SK blockers. Rapidly after birth, the outward current increases to 15nS-V and becomes sensitive to Ca2z, TEA, BK, and SK blockers. We make a detailed analysis of the kinetics of the components of the outward currents and produce a model for BK and SK channels that we use to reproduce the outward current, and to infer the geometrical arrangement of BK and Ca2z channels in clusters. In the first cluster, T-type Ca2z and BK channels are coupled within distances of *20 nm (200 A˚). The second cluster consists of L-type Ca2z and BK channels that are spread over distances of at least 60 nm. As for the second question, we propose that early in development, the mechanism of action selection is in a ‘‘locked-in’’ state that would prevent dopamine neurons from reinforcing cortico-striatal synapses that do not have a functional experiential- based value.

Functional Upregulation of Ca2+ -Activated K+ Channels in the Development of Substantia Nigra Dopamine Neurons

Many connections in the basal ganglia are made around birth when animals are exposed to a host of new affective, cognitive, and sensori-motor stimuli. It is thought that dopamine modulates cortico-striatal synapses that result in the strengthening of those connections that lead to desired outcomes. We propose that there must be a time before which stimuli cannot be processed into functional connections, otherwise it would imply an effective link between stimulus, response, and reward in uterus. Consistent with these ideas, we present evidence that early in development dopamine neurons are electrically immature and do not produce high-frequency firing in response to salient stimuli. We ask first, what makes dopamine neurons immature? and second, what are the implications of this immaturity for the basal ganglia? As an answer to the first question, we find that at birth the outward current is small (3nS-V), insensitive to Ca2+, TEA, BK, and SK blockers. Rapidly after birth, the outward current increases to 15nS-V and becomes sensitive to Ca2+, TEA, BK, and SK blockers. We make a detailed analysis of the kinetics of the components of the outward currents and produce a model for BK and SK channels that we use to reproduce the outward current, and to infer the geometrical arrangement of BK and Ca2+ channels in clusters. In the first cluster, T-type Ca2+ and BK channels are coupled within distances of similar to 20 nm (200 parallel to). The second cluster consists of L-type Ca2+ and BK channels that are spread over distances of at least 60 nm. As for the second question, we propose that early in development, the mechanism of action selection is in a "locked-in" state that would prevent dopamine neurons from reinforcing cortico-striatal synapses that do not have a functional experiential-based value.

Changes in heron and egret populations on the Laurentian Great Lakes and connecting channels, 1977-2009

Canadian and U.S. federal wildlife agencies completed four decadal surveys, spanning the years 1977 to 2009, to census colonial waterbirds breeding on the Great Lakes and adjoining bodies of water. In this paper, we reports abundance, distribution, and general population trends of three species: Black-crowned Night-Heron (Nycticorax nycticorax), Great Egret (Ardea alba), and Great Blue Heron (Ardea herodias). Estimates of nest numbers ranged from approximately 4000-6100 for the Black-crowned Night-Heron, 250-1900 for the Great Egret, and 3800-6400 for the Great Blue Heron. Average annual rates of change in nest numbers between the first (1977) and fourth (2008) census were −1% for the Black-crowned Night-Heron, +23% for the Great Egret, and −0.27% for the Great Blue Heron. Across the 30-year census, Black-crowned Night-Heron estimates decreased in U.S. (−57%) but increased (+18%) in Canadian waters, Great Egret nests increased 1381% in Canadian waters with a smaller, but still substantial increase in the number of nests at U.S. colonies (+613%), and Great Blue Heron numbers increased 148% in Canadian waters and 713% in U.S. waters. Although a single factor cannot be clearly linked to changes observed in each species’ distribution, hydrological variation, habitat succession, nest competition with Double-crested Cormorants (Phalacrocorax auritus), and land use changes likely all contributed. Management activities should support both breeding and foraging conditions including restoration of early successional habitats and anticipate continued northward expansions in the distributions of these waterbirds.