876 resultados para statistical learning mechanisms
Resumo:
We combine the replica approach from statistical physics with a variational approach to analyze learning curves analytically. We apply the method to Gaussian process regression. As a main result we derive approximative relations between empirical error measures, the generalization error and the posterior variance.
Resumo:
A novel approach, based on statistical mechanics, to analyze typical performance of optimum code-division multiple-access (CDMA) multiuser detectors is reviewed. A `black-box' view ot the basic CDMA channel is introduced, based on which the CDMA multiuser detection problem is regarded as a `learning-from-examples' problem of the `binary linear perceptron' in the neural network literature. Adopting Bayes framework, analysis of the performance of the optimum CDMA multiuser detectors is reduced to evaluation of the average of the cumulant generating function of a relevant posterior distribution. The evaluation of the average cumulant generating function is done, based on formal analogy with a similar calculation appearing in the spin glass theory in statistical mechanics, by making use of the replica method, a method developed in the spin glass theory.
Resumo:
Online learning is discussed from the viewpoint of Bayesian statistical inference. By replacing the true posterior distribution with a simpler parametric distribution, one can define an online algorithm by a repetition of two steps: An update of the approximate posterior, when a new example arrives, and an optimal projection into the parametric family. Choosing this family to be Gaussian, we show that the algorithm achieves asymptotic efficiency. An application to learning in single layer neural networks is given.
Resumo:
Fourier-phase information is important in determining the appearance of natural scenes, but the structure of natural-image phase spectra is highly complex and difficult to relate directly to human perceptual processes. This problem is addressed by extending previous investigations of human visual sensitivity to the randomisation and quantisation of Fourier phase in natural images. The salience of the image changes induced by these physical processes is shown to depend critically on the nature of the original phase spectrum of each image, and the processes of randomisation and quantisation are shown to be perceptually equivalent provided that they shift image phase components by the same average amount. These results are explained by assuming that the visual system is sensitive to those phase-domain image changes which also alter certain global higher-order image statistics. This assumption may be used to place constraints on the likely nature of cortical processing: mechanisms which correlate the outputs of a bank of relative-phase-sensitive units are found to be consistent with the patterns of sensitivity reported here.
Resumo:
Changes in the strength of signalling between neurones are thought to provide a cellular substrate for learning and memory. In the cerebellar cortex, raising the frequency and the strength of parallel fibre (PF) stimulation leads to a long-term depression (LTD) of the strength of signalling at the synapse between PFs and Purkinje cells (PCs), which spreads to distant synapses to the same cell via a nitric oxide (NO) dependent mechanism. At the same synapse, but under conditions of reduced post-synaptic calcium activity, raised frequency stimulation (RFS) of PFs triggers a long-term potentiation of synaptic transmission. The aims of the work described in this thesis were to investigate the conditions necessary for LTD and LTP at this synapse following RFS and to identify the origins and second messenger cascades involved in the induction and spread of LTP and LTD. In thin, parasagittal cerebellar slices whole cell patch clamp recordings were made from PCs and the effects of RFS of one of two, independent PF inputs to the same PC were examined under a range of experimental conditions. Under conditions designed to reduce post-synaptic calcium activity, RFS to a single PF input led to LTP and a decreases in paired pulse facilitation (PPF) in both pathways. This heterosynaptic potentiation was prevented by inhibition of protein kinase A (PKA) or by inhibition of NO synthase with either 7-nitroindazole (7-NI) or NG Nitro-L-argenine methyl ester. Inhibition of guanylate cyclase (GC) or protein kinase G (PKG) had no effect. A similar potentiation was observed upon application of the adenylyl cyclase (AC) activator forskolin or the NO donor spermine NONOate. Both of these treatments also resulted in an increase in the frequency of mEPSCs, which provides further evidence for a presynaptic origin of LTP. Forskolin induced potentiation and the increase in mEPSC frequency were blocked by 7-NI. The styryl dye FM1-43, a fluorescent reporter of endo- and exocytosis, was also used to further examine the possible pre-synaptic origins of LTP. RFS or forskolin application enhanced FM1-43 de-staining and NOS inhibitors blocked this effect. Application of NONOate also enhanced FM1-43 de-staining. When post-synaptic calcium activity was less strictly buffered, RFS to a single PF input led to a transient potentiation that was succeeded by LTD in both pathways. This LTD, which resembled previously described forms, was prevented by inhibition of the NO/cGMP/PKG cascade. Modification of the AC/cAMP/PKA cascade had no effect. In summary, the direction of synaptic plasticity at the PF-PC synapse in response to RFS depends largely on the level of post-synaptic calcium activity. LTP and LTD were non-input specific and both forms of plasticity were dependent on NOS activity. Induction of LTP was mediated by a presynaptic mechanism and depended on NO and cAMP production. LTD on the other hand was a post-synaptic process and required activity of the NO/cGMP/PKG signalling cascade.
Resumo:
Ten grades of ABS and four grades of polypropylene have been plated with various copper + nickel + chromium coatings and subjected to a variety of tests. In corrosion studies the pre-electroplating sequence and plastics type have been shown to influence performance. One ABS pre-electroplating sequence was consistently associated with better corrosion performance; two factors were responsible for this, namely the more severe nature of the etch and the relatively more noble electroless nickel. Statistical analysis has indicated that order of severity of the corrosion tests was static-mobile-CASS, the latter being the least severe. In mechanical tests two properties of ABS and polypropJylene, ductility and impact strength, have been shown to be adversely affected when electrodeposited layers were applied. The cause of this is due to a complex of factors, the most important of which is the notch sensitivity of the plastics. Peel adhesion has been studied on flat panels and also on ones which had a ridge and a valley moulded into one face. High adhesion peaks occurred on the flat face at regions associated with the ridge and valley. The local moulding conditions induced by the features were responsible for this phenonemon. In the main programme the thermal cycling test was shown to be more likely than the peel adhesion test to give an indication of the service performance of electroplated plastics.
Resumo:
Background - The literature is not univocal about the effects of Peer Review (PR) within the context of constructivist learning. Due to the predominant focus on using PR as an assessment tool, rather than a constructivist learning activity, and because most studies implicitly assume that the benefits of PR are limited to the reviewee, little is known about the effects upon students who are required to review their peers. Much of the theoretical debate in the literature is focused on explaining how and why constructivist learning is beneficial. At the same time these discussions are marked by an underlying presupposition of a causal relationship between reviewing and deep learning. Objectives - The purpose of the study is to investigate whether the writing of PR feedback causes students to benefit in terms of: perceived utility about statistics, actual use of statistics, better understanding of statistical concepts and associated methods, changed attitudes towards market risks, and outcomes of decisions that were made. Methods - We conducted a randomized experiment, assigning students randomly to receive PR or non–PR treatments and used two cohorts with a different time span. The paper discusses the experimental design and all the software components that we used to support the learning process: Reproducible Computing technology which allows students to reproduce or re–use statistical results from peers, Collaborative PR, and an AI–enhanced Stock Market Engine. Results - The results establish that the writing of PR feedback messages causes students to experience benefits in terms of Behavior, Non–Rote Learning, and Attitudes, provided the sequence of PR activities are maintained for a period that is sufficiently long.
Resumo:
The problem of learning by examples in ultrametric committee machines (UCMs) is studied within the framework of statistical mechanics. Using the replica formalism we calculate the average generalization error in UCMs with L hidden layers and for a large enough number of units. In most of the regimes studied we find that the generalization error, as a function of the number of examples presented, develops a discontinuous drop at a critical value of the load parameter. We also find that when L>1 a number of teacher networks with the same number of hidden layers and different overlaps induce learning processes with the same critical points.
Resumo:
In the editorial of this special issue we argue that knowledge flows, learning and development are becoming increasingly important in all organisations operating in an international context. The possession of capabilities relating to acquisition, configuration and transfer of relevant knowledge effectively within and across different organisational units, teams, and countries is integrally related to superior organisational performance. In mastering such capabilities, internationalised organisations need to grapple with the inherent challenges relating to contextual variation and different work modes between subsidiaries, partners or team members. The papers in this special issue cast light on crucial aspects of knowledge flows, learning and development in internationalised organisations. Their contribution varies from the provision of frameworks to systematise investigation of these issues, to empirical evidence about effective mechanisms, as well as enabling and constraining forces, in facilitating knowledge transfer, learning and human capital development. © 2012 Inderscience Enterprises Ltd.
Resumo:
Although the role of strategic leadership in developing organizational learning capabilities has been examined to a certain extent, the specific behaviors and mechanisms through which these capabilities are developed have not been adequately understood. Paucity of research in this direction is even more conspicuous in a small firm/entrepreneurship context, which has been linked to innovation, economic growth, and employment generation. Reporting on an ethnographic study of a knowledge-intensive, growth-oriented small firm, this article addresses this gap in the literature by integrating strategic leadership and organizational learning theory in an entrepreneurship context. In this undertaking, situated learning theory is used as the major analytical lens, to shed light on how strategic leadership can build organizational learning capabilities that underpin entrepreneurial performance in small firms. Finally, implications for situated learning theory as an organizational learning perspective and leadership practice in an entrepreneurship context are submitted. © 2012 Wiley Periodicals, Inc.
Resumo:
In part 1 of this article, cleavage initiation in the intercritically reheated coarse-grained heat affected zone (IC CG HAZ) of high-strength low-alloy (HSLA) steels was determined to occur between two closely spaced blocky MA particles. Blunt notch, crack tip opening displacement (CTOD), and precracked Charpy testing were used in this investigation to determine the failure criteria required for cleavage initiation to occur by this mechanism in the IC CG HAZ. It was found that the attainment of a critical level of strain was required in addition to a critical level of stress. This does not occur in the case of high strain rate testing, for example, during precracked Charpy testing. A different cleavage initiation mechanism is then found to operate. The precise fracture criteria and microstructural requirements (described in part I of this article) result in competition between potential cleavage initiation mechanisms in the IC CG HAZ.
Resumo:
Four bar mechanisms are basic components of many important mechanical devices. The kinematic synthesis of four bar mechanisms is a difficult design problem. A novel method that combines the genetic programming and decision tree learning methods is presented. We give a structural description for the class of mechanisms that produce desired coupler curves. Constructive induction is used to find and characterize feasible regions of the design space. Decision trees constitute the learning engine, and the new features are created by genetic programming.
Resumo:
Spatial objects may not only be perceived visually but also by touch. We report recent experiments investigating to what extent prior object knowledge acquired in either the haptic or visual sensory modality transfers to a subsequent visual learning task. Results indicate that even mental object representations learnt in one sensory modality may attain a multi-modal quality. These findings seem incompatible with picture-based reasoning schemas but leave open the possibility of modality-specific reasoning mechanisms.
Resumo:
E-learning means learning via electronic means and is therefore an all-embracing term covering learning via an electronic device. The "expectations" and "realities" for each of the delivery mechanisms within the electronic arena vary greatly for not just the learners themselves, but also the site providers. Because of this, each of these learning systems has vastly different design principles, which is not always understood by those unfamiliar with technology. What is appropriate for a CD-ROM off-line system is generally inappropriate for an on- line internet system. So when designing an e-learning system it is important to understand how the information is to be accessed by the learner. This paper will identify and suggest some ways to avoid e-learning's pitfalls and reap its rewards.
Resumo:
When Recurrent Neural Networks (RNN) are going to be used as Pattern Recognition systems, the problem to be considered is how to impose prescribed prototype vectors ξ^1,ξ^2,...,ξ^p as fixed points. The synaptic matrix W should be interpreted as a sort of sign correlation matrix of the prototypes, In the classical approach. The weak point in this approach, comes from the fact that it does not have the appropriate tools to deal efficiently with the correlation between the state vectors and the prototype vectors The capacity of the net is very poor because one can only know if one given vector is adequately correlated with the prototypes or not and we are not able to know what its exact correlation degree. The interest of our approach lies precisely in the fact that it provides these tools. In this paper, a geometrical vision of the dynamic of states is explained. A fixed point is viewed as a point in the Euclidean plane R2. The retrieving procedure is analyzed trough statistical frequency distribution of the prototypes. The capacity of the net is improved and the spurious states are reduced. In order to clarify and corroborate the theoretical results, together with the formal theory, an application is presented