Statistical mechanical analysis of the dynamics of learning in perceptrons

We describe the application of tools from statistical mechanics to analyse the dynamics of various classes of supervised learning rules in perceptrons. The character of this paper is mostly that of a cross between a biased non-encyclopaedic review and lecture notes: we try to present a coherent and self-contained picture of the basics of this field, to explain the ideas and tricks, to show how the predictions of the theory compare with (simulation) experiments, and to bring together scattered results. Technical details are given explicitly in an appendix. In order to avoid distraction we concentrate the references in a final section. In addition this paper contains some new results: (i) explicit solutions of the macroscopic equations that describe the error evolution for on-line and batch learning rules; (ii) an analysis of the dynamics of arbitrary macroscopic observables (for complete and incomplete training sets), leading to a general Fokker–Planck equation; and (iii) the macroscopic laws describing batch learning with complete training sets. We close the paper with a preliminary expose´ of ongoing research on the dynamics of learning for the case where the training set is incomplete (i.e. where the number of examples scales linearly with the network size).     

A Full Bayesian Non-parametric Analysis Involving a Neutral to the Right Process

Implementation of a full Bayesian non-parametric analysis involving neutral to the right processes (apart from the special case of the Dirichlet process) has been difficult for two reasons: first, the posterior distributions are complex and therefore only Bayes estimates (posterior expectations) have previously been presented; secondly, it is difficult to obtain an interpretation for the parameters of a neutral to the right process. In this paper we extend Ferguson & Phadia (1979) by presenting a general method for specifying the prior mean and variance of a neutral to the right process, providing the interpretation of the parameters. Additionally, we provide the basis for a full Bayesian analysis, via simulation, from the posterior process using a hybrid of new algorithms that is applicable to a large class of neutral to the right processes (Ferguson & Phadia only provide posterior means). The ideas are exemplified through illustrative analyses.     

Are We Playing the Same Game?: Publishing Task Environments and Research Productivity Among Management Specialists

Archival data on over 15,000 contributors to 33management journals are used to examine productivitydifferences among scholars in different areas ofmanagement. The results suggest that a fragmentedpublication network offers specialists from different areasdissimilar publication opportunities and experiences,and may contribute to varying levels of publicationproductivity among authors. The implications of our results for future research and facultyevaluation processes are explored.     

Do Status Differences among Workers Make a Difference during Economic Crises? The Case of Depression Hamilton*

Utilisant les données du recensement de 1931 ainsi que d'entrevues avec des retraités d'Hamilton, en Ontario, qui recevaient un salaire durant la Grande Dépression, l'auteur examine les affirmations selon lesquelles les crises économiques sapent ou renforcent, ou n'affectent pas nécessairement la hiérarchie traditionnelle. Cela parce que, respectivement, elles substituent des ouvriers moins qualifiés, à meilleur marché aux autres ouvriers, elles privilégient les travailleurs hautement qualifiés ou elles affectent différemment les industries et occupations parmi lesquelles les ouvriers de différentes conditions sociales sont ségrégués. Le plus notable est qu'aucun résultat n'était marqué ni consistant en ce qui concerne les situations professionnelles, lâge, le sexe et l'ethnie sur tout type de privation. Cependant, certains types de privation montraient certaines différences de rangs. Donc, lorsque la profession était contrôlée dans les données du recensement, le modéle le plus commun était que les femmes travaillaient plus que les hommes, seulement lorsqu'en fait elles gagnaient moins. Cela suggère la «substitution» du travail à bon marché des femmes à celui des hommes et, en conséquence, l'ébranlement des privilèges de ces derniers. Par contre, selon les données tirées des entrevues, plus on était jeune et plus on attendait pour obtenir un travail rémunéréà temps plein, plus il était difficile d'en trouver, et plus le salaire était bas. Cela suggère que des privilèges associés avec lâge étaient renforcés durant la Dépression. Néanmoins, il y avait aussi ségrégation industrielle et occupationnelle, selon lâge, l'ethnie et le sexe, ce qui peut expliquer en partie pourquoi la plupart des différences de rangs sociaux n'étaient ni sapées ni renforcées. L'auteur a aussi trouvé peu de différences de rangs sociaux dans les réactions des ouvriers à la crise et, par conséquent, peu d'évidences quant au processus d'homogénéisation/radicalisation de la gauche associé avec l'hypothèse de la première affirmation ou à la diffé-renciation/conservatisme des ouvriers hautement qualifiés et à la radicalisation des ouvriers moins qualifiés associés avec l'affir-mation de l'hypothèse de renforcissement. Using data from the 1931 Census and interviews with Hamilton, Ontario retirees who worked for pay during the Great Depression, the author examines claims that economic crises undermine, strengthen, or need not affect traditional status hierarchies; because, respectively, they substitute “cheaper” low-status workers for others; privilege high-status workers; or differentially affect industries and occupations where workers of different status are segregated. The most remarkable result was that there were no such strong and consistent results for any of occupational status, age, gender and ethnicity on all types of deprivation. However, there was evidence for some status differences on some types of deprivation. Thus, when occupation was controlled in the census data, the most common pattern was for women to work more than men only when they in fact earned less. This suggests the “substitution” of the cheaper labour of women for that of men, and therefore the undermining of the privileges of men. On the other hand, in the interview data, the younger one was and the later one attempted to obtain full-time, paid work, the more difficulty one had finding it, and the lower one's wages were. This suggests that some age privileges were strengthened during the Depression. Nevertheless, there was also much industrial and occupational segregation, by age and ethnicity as well as gender, and this may partly explain why most status differences were neither undermined nor strengthened. The author also found few status differences in workers' responses to the crisis, and therefore little evidence for either the homogenization/left-radicalization process associated with the undermining hypothesis or the differentiation/conservatizing of high-status workers and radicalization of low-status workers associated with the strengthening claim.     

PRODUCT CHARACTERISTICS, HUMAN CAPITAL, AND LABOR MARKET SEGMENTS:

This article contends that one key to understanding different forms of work organization lies in the nature of the products being created. Product characteristics are proposed to be critical determinants of the type of human capital, either general or firm specific. Following from prior theory, labor market barriers develop based on type of human capital. These barriers then have a direct bearing on employee rewards. The nature of the product distinction is captured with a comparison of two product-types (goods and services) conceived as theoretically distinct. General skills are hypothesized to be more important in the service-producing sector, while firm-specific skills are hypothesized to be more important in the goods-producing sector. Empirical analyses using the 1991 General Social Survey compare workers in the service-producing and goods-producing sectors to illustrate differences in the salience of firm-specific and general skills. Two hypotheses are supported. Firm-specific skills have a stronger effect on earnings in manufacturing industries than in service industries. Also, skills acquired from on-the-job training, when compared with other skills, are more weakly related to service employee rewards. These distinctions between sectors suggest insights into structures unique to the service employment workplace.     

Information subsidies and agenda-building during the Israel–Lebanon crisis

This study examined the impact of information subsidies on media coverage during a crisis. Using the July 2006 Israel–Lebanon conflict as a backdrop, this research reviewed access that U.S. military public affairs officers provided the media and analyzed subsequent coverage for the presence of the military's message. Coverage was more neutral to positive than negative. Items containing organizational messages were more positive; those quoting practitioner-facilitated sources introduced organizational messages into coverage and generated more positive coverage. Access to information subsidies had a positive impact on coverage and aided in the successful transfer of attribute salience from practitioners to the media.     

Trust in surveys and games – A methodological contribution on the influence of money and location

This paper explores methods to study trust. In a variety of settings, answers to survey questions and choices in a trust game are obtained from student sample pools. Some subjects are approached by mail and execute their task at home whereas others participate in classroom experiments. No differences between the results obtained by these methods are observed. Furthermore, one additional group plays the trust game with purely hypothetical payments, and another receives random lottery payments. This changes trust behavior dramatically, whereas trustworthiness is unaffected. Subjects without any financial incentives exhibit less trust and their trust choices are significantly correlated with survey trust answers. There is no such correlation for the corresponding choices with real payments.     

A critique of arguments supporting disability avoidance

Many prominent bioethicists argue that positive consequences follow from avoiding disability in offspring by preconceptual, preimplantation or prenatal means. This paper critiques two variants of these positions: ‘Open future’ arguments that avoiding disability is in the interest of the persons who will be affected by it if born and ‘non‐person‐affecting’ claims that failures to avoid disability in offspring are wrong for reasons independent of their effect on affected offspring. Difficulties with both person‐affecting and non‐person‐affecting accounts are discussed in support of the charge that these arguments do not show that positive consequences follow from disability avoidance.