Noise, regularizers, and unrealizable scenarios in online learning from restricted training sets

We study the dynamics of on-line learning in multilayer neural networks where training examples are sampled with repetition and where the number of examples scales with the number of network weights. The analysis is carried out using the dynamical replica method aimed at obtaining a closed set of coupled equations for a set of macroscopic variables from which both training and generalization errors can be calculated. We focus on scenarios whereby training examples are corrupted by additive Gaussian output noise and regularizers are introduced to improve the network performance. The dependence of the dynamics on the noise level, with and without regularizers, is examined, as well as that of the asymptotic values obtained for both training and generalization errors. We also demonstrate the ability of the method to approximate the learning dynamics in structurally unrealizable scenarios. The theoretical results show good agreement with those obtained by computer simulations.

Binocular interaction:Contrast matching and contrast discrimination are predicted by the same model

How do signals from the 2 eyes combine and interact? Our recent work has challenged earlier schemes in which monocular contrast signals are subject to square-law transduction followed by summation across eyes and binocular gain control. Much more successful was a new 'two-stage' model in which the initial transducer was almost linear and contrast gain control occurred both pre- and post-binocular summation. Here we extend that work by: (i) exploring the two-dimensional stimulus space (defined by left- and right-eye contrasts) more thoroughly, and (ii) performing contrast discrimination and contrast matching tasks for the same stimuli. Twenty-five base-stimuli made from 1 c/deg patches of horizontal grating, were defined by the factorial combination of 5 contrasts for the left eye (0.3-32%) with five contrasts for the right eye (0.3-32%). Other than in contrast, the gratings in the two eyes were identical. In a 2IFC discrimination task, the base-stimuli were masks (pedestals), where the contrast increment was presented to one eye only. In a matching task, the base-stimuli were standards to which observers matched the contrast of either a monocular or binocular test grating. In the model, discrimination depends on the local gradient of the observer's internal contrast-response function, while matching equates the magnitude (rather than gradient) of response to the test and standard. With all model parameters fixed by previous work, the two-stage model successfully predicted both the discrimination and the matching data and was much more successful than linear or quadratic binocular summation models. These results show that performance measures and perception (contrast discrimination and contrast matching) can be understood in the same theoretical framework for binocular contrast vision. © 2007 VSP.

Bayesian online algorithms for learning in discrete Hidden Markov Models

We propose and analyze two different Bayesian online algorithms for learning in discrete Hidden Markov Models and compare their performance with the already known Baldi-Chauvin Algorithm. Using the Kullback-Leibler divergence as a measure of generalization we draw learning curves in simplified situations for these algorithms and compare their performances.

Utilising online shopping environment attributes holistically to create competitive advantage

As more consumers shop online, it becomes crucial for marketers to know how online shopping environments (OSEs) can be used to gain competitive advantage. This dissertation aims to explain theoretically how OSE attributes work together holistically to produce desirable consumer responses, applying and extending a theory from the environmental psychology literature to the online context. Firstly, the study conceptualises OSEs as virtual environments which may be perceived and experienced both cognitively and affectively through a technology-mediated interaction with a computer screen. A multi-disciplinary approach identifies key characteristics of OSEs: they involve consumers; they are more complex than their offline counterparts; they are likely first apprehended holistically; and they can elicit high levels of emotions and cognition. Secondly, the research uses a gestalt approach and extends Kaplan and Kalan’s (1982) Preference Framework, taking account of the specific characteristics of OSEs, which one visits specifically to obtain product information. The results support the proposition that OSEs are perceived in terms of their Sense-making and Exploratory attributes. Thirdly, the research explains how OSE attributes work together to produce desirable consumer responses. As hypothesised, Exploratory potential produces both Hedonic and Utilitarian value, and both kinds of value contribute to Site commitment. An unexpected result is that Sense-making potential does not produce Utilitarian value directly, but only through the mediation of Exploratory potential. The research contributes to marketing theory by: (1) identifying ways the internet has changed the nature of the shopping experience; (2) extending Kaplan and Kaplan’s Preference Framework to explain how consumers perceive OSEs holistically; (3) identifying the distinction between page-level and site-level perceptions, and (4) distinguishing between different sources of information (marketer vs. non-marketer). Managerially, the research provides a model for marketers to conceive and design retail websites whose attributes work together to create competitive advantage.

Interaction of quinolone antibiotics with the human intestinal CACO-2 cell model

The transport of a group of quinolone antibiotics across the human intestinal model, Caco-2 cells, was investigated. It was found that the transport of the quinolones generally correlated with the lipophilicity of the compounds, indicating the passive diffusional transcellular processes were involved. However, it was observed that the transport in both directions apical-to-basolateral and basolateral-to-apical was not equivalent, and polarised transport occurred. For all the quinolones studied except, BMS-284756-01, it was found that the basolateral-to-apical transport was significantly greater than the apical-to-basolateral transport. This finding suggested that the quinolones underwent a process of active secretion. The pKas and logPs for the quinolones were determined using potentiometric titrations. The measured logP values were compared with those determined using theoretical methods. The theoretical methods for calculating logP including the Moriguchi method correlated poorly with the measured logP values. Further investigations revealed that there may be an active transporter involved in the apical-to-basolateral transport of quinolones as well. This mechanism was sensitive to competing quinolones, but, it was unaffected by the metabolic inhibitor combination of sodium azide (15mM) with 2-deoxy-D-glucose (50mM). The basolateral-to-apical transport of quinolones was found to be sensitive to inhibition by a number of different inhibitors. The metabolic inhibitors, sodium azide (15mM) with 2-deoxy-D-glucose (50mM) and 2,4-dinitrophenol (1mM), were able to reduce the basolateral-to-apical transport of quinolones. A reduction in temperature from 37°C to 2°C caused an 80-fold decrease in the transport of gatifloxacin in both directions, however, this effect was not sufficient to abolish the greater basolateral-to-apical secretion. As with apical-to-basolateral transport, it was found that quinolones competed with gatifloxacin for basolateral-to-apical transport, both ofloxacin (100μM) and norfloxacin (100μM) significantly (P<0.003) decreased the basolateral-to-apical transport of gatifloxacin; however, ciprofloxacin (100μM and 300μM) had no effect. A number of inhibitors of various transport systems were also investigated. It was found that the anion transport inhibitor, probenecid (100 μM) had a significant inhibitory effect on the basolateral-to-apical transport of ciprofloxacin (P=0.039), while the cation transport inhibitor cimetidine (100μM and 500μM) had no effect. The organic anion exchange inhibitor 4,4'diisothiocyanostilbene-2-2' -disulphonic acid DIDS (400μM) also had a significant inhibitory effect (P=O.O 13). The PgP inhibitor and anion exchange inhibitor verapamil (400Mμ) was able to completely abolish the basolateral-to-apical secretion of gatifloxacin and bring it into line with the apical-to-basolateral flux. In conclusion, the apical-to-basolateral and basolateral-toapical transport of quinolones involved an active component. The basolateral-to-apical secretion was abolished by a verapamil (400μM), a bisubstrate for PgP and the anion transporter.

Learning sentiment classification model from labeled features

We propose a novel framework where an initial classifier is learned by incorporating prior information extracted from an existing sentiment lexicon. Preferences on expectations of sentiment labels of those lexicon words are expressed using generalized expectation criteria. Documents classified with high confidence are then used as pseudo-labeled examples for automatical domain-specific feature acquisition. The word-class distributions of such self-learned features are estimated from the pseudo-labeled examples and are used to train another classifier by constraining the model's predictions on unlabeled instances. Experiments on both the movie review data and the multi-domain sentiment dataset show that our approach attains comparable or better performance than exiting weakly-supervised sentiment classification methods despite using no labeled documents.

A generative model for online depth fusion

We present a probabilistic, online, depth map fusion framework, whose generative model for the sensor measurement process accurately incorporates both long-range visibility constraints and a spatially varying, probabilistic outlier model. In addition, we propose an inference algorithm that updates the state variables of this model in linear time each frame. Our detailed evaluation compares our approach against several others, demonstrating and explaining the improvements that this model offers, as well as highlighting a problem with all current methods: systemic bias. © 2012 Springer-Verlag.

Is there a global model of learning organizations? An empirical, cross-nation study

This paper develops and tests a learning organization model derived from HRM and dynamic capability literatures in order to ascertain the model's applicability across divergent global contexts. We define a learning organization as one capable of achieving on-going strategic renewal, arguing based on dynamic capability theory that the model has three necessary antecedents: HRM focus, developmental orientation and customer-facing remit. Drawing on a sample comprising nearly 6000 organizations across 15 countries, we show that learning organizations exhibit higher performance than their less learning-inclined counterparts. We also demonstrate that innovation fully mediates the relationship between our conceptualization of the learning organization and organizational performance in 11 of the 15 countries we examined. It is the first time in our knowledge that these questions have been tested in a major, cross-global study, and our work contributes to both HRM and dynamic capability literatures, especially where the focus is the applicability of best practice parameters across national boundaries.

Learning from openness : the dynamics of breadth in external innovation linkages

We explore how openness in terms of external linkages generates learning effects, which enable firms to generate more innovation outputs from any given breadth of external linkages. Openness to external knowledge sources, whether through search activity or linkages to external partners in new product development, involves a process of interaction and information processing. Such activities are likely to be subject to a learning process, as firms learn which knowledge sources and collaborative linkages are most useful to their particular needs, and which partnerships are most effective in delivering innovation performance. Using panel data from Irish manufacturing plants, we find evidence of such learning effects: establishments with substantial experience of external collaborations in previous periods derive more innovation output from openness in the current period. © 2013 The Authors. Strategic Management Journal published by John Wiley & Sons Ltd.

A four-factor user interaction model for content-based image retrieval

In order to bridge the “Semantic gap”, a number of relevance feedback (RF) mechanisms have been applied to content-based image retrieval (CBIR). However current RF techniques in most existing CBIR systems still lack satisfactory user interaction although some work has been done to improve the interaction as well as the search accuracy. In this paper, we propose a four-factor user interaction model and investigate its effects on CBIR by an empirical evaluation. Whilst the model was developed for our research purposes, we believe the model could be adapted to any content-based search system.

The development of a computer simulation model of the interaction between the supply and usage of major components in car assembly you use in stock control

DUE TO COPYRIGHT RESTRICTIONS ONLY AVAILABLE FOR CONSULTATION AT ASTON UNIVERSITY LIBRARY AND INFORMATION SERVICES WITH PRIOR ARRANGEMENT