Measurements of generalisation based on information geometry

Neural networks are statistical models and learning rules are estimators. In this paper a theory for measuring generalisation is developed by combining Bayesian decision theory with information geometry. The performance of an estimator is measured by the information divergence between the true distribution and the estimate, averaged over the Bayesian posterior. This unifies the majority of error measures currently in use. The optimal estimators also reveal some intricate interrelationships among information geometry, Banach spaces and sufficient statistics.

Bayesian invariant measurements of generalisation

The problem of evaluating different learning rules and other statistical estimators is analysed. A new general theory of statistical inference is developed by combining Bayesian decision theory with information geometry. It is coherent and invariant. For each sample a unique ideal estimate exists and is given by an average over the posterior. An optimal estimate within a model is given by a projection of the ideal estimate. The ideal estimate is a sufficient statistic of the posterior, so practical learning rules are functions of the ideal estimator. If the sole purpose of learning is to extract information from the data, the learning rule must also approximate the ideal estimator. This framework is applicable to both Bayesian and non-Bayesian methods, with arbitrary statistical models, and to supervised, unsupervised and reinforcement learning schemes.

Distributed control of computer communication systems

Flow control in Computer Communication systems is generally a multi-layered structure, consisting of several mechanisms operating independently at different levels. Evaluation of the performance of networks in which different flow control mechanisms act simultaneously is an important area of research, and is examined in depth in this thesis. This thesis presents the modelling of a finite resource computer communication network equipped with three levels of flow control, based on closed queueing network theory. The flow control mechanisms considered are: end-to-end control of virtual circuits, network access control of external messages at the entry nodes and the hop level control between nodes. The model is solved by a heuristic technique, based on an equivalent reduced network and the heuristic extensions to the mean value analysis algorithm. The method has significant computational advantages, and overcomes the limitations of the exact methods. It can be used to solve large network models with finite buffers and many virtual circuits. The model and its heuristic solution are validated by simulation. The interaction between the three levels of flow control are investigated. A queueing model is developed for the admission delay on virtual circuits with end-to-end control, in which messages arrive from independent Poisson sources. The selection of optimum window limit is considered. Several advanced network access schemes are postulated to improve the network performance as well as that of selected traffic streams, and numerical results are presented. A model for the dynamic control of input traffic is developed. Based on Markov decision theory, an optimal control policy is formulated. Numerical results are given and throughput-delay performance is shown to be better with dynamic control than with static control.

A Kullback-Leibler relat��v entr��pia f��ggv��ny alkalmaz��sa p��ros ��sszehasonl��t��s m��trix egy priorit��svektora meghat��roz��s��ra (Applying the Kullback-Leibler relative entropy function for determining priorities for the pairwise comparison matrix)

A dolgozatban a d��nt��selm��letben fontos szerepet j��tsz�� p��ros ��sszehasonl��t��s m��trix priorit��svektor��nak meghat��roz��s��ra ��j megk��zel��t��st alkalmazunk. Az A p��ros ��sszehasonl��t��s m��trix ��s a priorit��svektor ��ltal defini��lt B konzisztens m��trix k��z��tti elt��r��st a Kullback-Leibler relat��v entr��pia-f��ggv��ny seg��ts��g��vel m��rj��k. Ezen elt��r��s minimaliz��l��sa teljesen kit��lt��tt m��trix eset��ben konvex programoz��si feladathoz vezet, nem teljesen kit��lt��tt m��trix eset��ben pedig egy fixpont probl��m��hoz. Az elt��r��sf��ggv��nyt minimaliz��l�� priorit��svektor egyben azzal a tulajdons��ggal is rendelkezik, hogy az A m��trix elemeinek ��sszege ��s a B m��trix elemeinek ��sszege k��z��tti k��l��nbs��g ��ppen az elt��r��sf��ggv��ny minimum��nak az n-szerese, ahol n a feladat m��rete. ��gy az elt��r��sf��ggv��ny minimum��nak ��rt��ke k��t szempontb��l is lehet alkalmas az A m��trix inkonzisztenci��j��nak a m��r��s��re. _____ In this paper we apply a new approach for determining a priority vector for the pairwise comparison matrix which plays an important role in Decision Theory. The divergence between the pairwise comparison matrix A and the consistent matrix B defined by the priority vector is measured with the help of the Kullback-Leibler relative entropy function. The minimization of this divergence leads to a convex program in case of a complete matrix, leads to a fixed-point problem in case of an incomplete matrix. The priority vector minimizing the divergence also has the property that the difference of the sums of elements of the matrix A and the matrix B is n times the minimum of the divergence function where n is the dimension of the problem. Thus we developed two reasons for considering the value of the minimum of the divergence as a measure of inconsistency of the matrix A.

Relaci��n entre los riesgos percibidos por los trabajadores en miner��a subterr��nea y los accidentes de trabajo y enfermedades laborales

Introducci��n: La miner��a es considerada uno de los sectores econ��micos m��s importantes por su capacidad para generar recursos en su propio sector y en otros sectores como metalmec��nica, agricultura e inform��tica entre otros, y por su contribuci��n al desarrollo socioecon��mico sostenible de las poblaciones. Objetivo: Determinar la relaci��n entre los riesgos percibidos por los trabajadores que laboran en miner��a subterr��nea en 3 departamentos de Colombia y los Accidentes de Trabajo (AT) y Enfermedades Laborales (EL). Materiales y M��todos: Estudio de corte transversal en 476 trabajadores de miner��a subterr��nea. Se incluyeron variables independientes (caracter��sticas sociodemogr��ficas y laborales y percepci��n del riesgo) y variables dependientes (enfermedad laboral y accidente de trabajo), obtenidas a trav��s de una entrevista directa aplicada por profesionales de la salud previamente capacitados. Para el an��lisis estad��stico se utiliz�� la Prueba Exacta de Fisher, el Odds Ratio (OR) con el Intervalo de Confianza (IC) del 95%. Resultados: En los trabajadores de miner��a subterr��nea en los departamentos de Boyac��, Cundinamarca y Santander, se encontr�� relaci��n estad��stica significativa entre la accidentalidad con la percepci��n de riesgo por iluminaci��n (OR= 2.059, IC= 95%: 1.116, 3.798, p=0.013), percepci��n de riesgo por movimientos repetitivos (OR= 1.951, IC= 95%: 0.998, 3.815, p=0.034), percepci��n de riesgo por ruido (OR= 2.275, IC= 95%: 0.974, 5.312, p=0.039) y percepci��n de riesgo por manejo de cargas (OR= 1.778, IC= 95%: 0.969, 3.264, p=0.041). Conclusi��n: se encontr�� que existe una relaci��n significativa entre la percepci��n de riesgo de los trabajadores de miner��a subterr��nea con accidentes de trabajo y que no existe relaci��n entre esta percepci��n y las enfermedades laborales.

la influencia de la felicidad en las decisiones de inversi��n de activos de largo plazo para una muestra de siete gerentes financieros en la ciudad de Bogot�� en el a��o 2016.

El presente art��culo contribuye con la investigaci��n de las Finanzas Corporativas del Comportamiento, rama de las finanzas corporativas que considera que el individuo que toma decisiones financieras no es completamente racional y que por hecho existen sesgos psicol��gicos que influyen en sus decisiones. Este documento se enfoca, desde el punto de vista conceptual y tambi��n mediante el an��lisis de un estudio de campo, en la influencia de la felicidad en las decisiones de inversi��n en activos de largo plazo para un grupo de siete gerentes ubicados en la ciudad de Bogot�� en el a��o 2016. En el documento se abarca el concepto general de las finanzas corporativas del comportamiento, se define la felicidad y se presentan sub-variables determinantes para la felicidad del individuo como lo son: salud, balance vida/trabajo, educaci��n y habilidades, conexiones sociales y medio ambiente. Finalmente se presenta c��mo ��stas afectan a los gerentes financieros en sus decisiones de acuerdo a la investigaci��n realizada.

Volunteer decision making by older people: A test of a revised theory of planned behavior

This study was designed to test the utility of a revised theory of planned behavior in the prediction of intentions to volunteer among older people. Such a perspective allowed for the consideration of a broader range of social and contextual factors than has been examined in previous research on volunteer decision making among older people. The article reports the findings from a study that investigated volunteer intentions and behavior in a random sample of older people aged 65 to 74 years living in an Australian capital city. Results showed that, as predicted by the revised theory of planned behavior, intention to volunteer predicted subsequent reported volunteer behavior. Intention was, in turn, predicted by social norms (both subjective and behavioral), perceived behavioral control, and moral obligation, with the effect of attitude being mediated through moral obligation.

Decision Support for Energy Contracts Negotiation with Game Theory and Adaptive Learning

This paper presents a decision support methodology for electricity market players��� bilateral contract negotiations. The proposed model is based on the application of game theory, using artificial intelligence to enhance decision support method���s adaptive features. This model is integrated in AiD-EM (Adaptive Decision Support for Electricity Markets Negotiations), a multi-agent system that provides electricity market players with strategic behavior capabilities to improve their outcomes from energy contracts��� negotiations. Although a diversity of tools that enable the study and simulation of electricity markets has emerged during the past few years, these are mostly directed to the analysis of market models and power systems��� technical constraints, making them suitable tools to support decisions of market operators and regulators. However, the equally important support of market negotiating players��� decisions is being highly neglected. The proposed model contributes to overcome the existing gap concerning effective and realistic decision support for electricity market negotiating entities. The proposed method is validated by realistic electricity market simulations using real data from the Iberian market operator���MIBEL. Results show that the proposed adaptive decision support features enable electricity market players to improve their outcomes from bilateral contracts��� negotiations.

A new aggregation method for strategic decision making and its application in assignment theory

A new aggregation method for decision making is presented by using induced aggregation operators and the index of maximum and minimum level. Its main advantage is that it can assess complex reordering processes in the aggregation that represent complex attitudinal characters of the decision maker such as psychological or personal factors. A wide range of properties and particular cases of this new approach are studied. A further generalization by using hybrid averages and immediate weights is also presented. The key issue in this approach against the previous model is that we can use the weighted average and the ordered weighted average in the same formulation. Thus, we are able to consider the subjective attitude and the degree of optimism of the decision maker in the decision process. The paper ends with an application in a decision making problem based on the use of the assignment theory.

Factors affecting a woman's decision to leave an abusive relationship : the theory of planned behaviour /

The present study tested the appHcabiUty of Ajzen's (1985) theory of planned behaviour (TPB), an extension of Fishbein and Ajzen's (1975) theory of reasoned action (TRA), for the first time, in the context of abused women's decision to leave their abusive relationships. The TPB, as a means of predicting women's decision to leave their abusive partners' was drawn from Strube's (1988, 1991) proposed decision-making model based on the principle that the decision-making process is a rational, deliberative process, and regardless of outcome, was a result of a logical assessment of the available data. As a means of predicting those behaviours not under volitional control, Ajzen's (1985) TPB incorporated a measure of perceived behavioural control. Data were collected in two phases, ranging from 6 months to 1 year apart. It was hypothesized that, to the extent that an abused woman held positive attitudes, subjective norms conducive to leaving, and perceived control over leaving, she would form an intention to leave and thus, increase the likelihood of actually leaving her partner. Furthermore, it was expected that perceptions of control would predict leaving behaviour over and above attitude and subjective norm. In addition, severity and frequency of abuse were assessed, as were demographic variables. The TPB failed to account significantly for variability in either intentions or leaving behaviour. All of the variance was attributed to those variables associated with the theory of reasoned action, with social influence emerging as the strongest predictor of a woman's intentions. The poor performance of this model is attributed to measurement problems with aspects of attitude and perceived control, as well as a lack of power due to the small sample size. The insufficiency of perceived control to predict behaviour also suggests that, on the surface at least, other factors may be at work in this context. Implications of these results, and recommendations such as, the importance of obtaining representative samples, the inclusion of self-esteem and emotions as predictor variables in this model, a reevaluation of the target behaviovu" as nonvolitional, and longitudinal studies spanning a longer time period for future research within the context of decision-making are discussed.

Decision-making theory applied to India's explosion of a nuclear device in May, 1974

This essay reviews the decision-making process that led to India exploding a nuclear device in May, 1974. An examination of the Analytic, Cybernetic and Cognitive Theories of decision, will enable a greater understanding of the events that led up to the 1974 test. While each theory is seen to be only partially useful, it is only by synthesising the three theories that a comprehensive account of the 1974 test can be given. To achieve this analysis, literature on decision-making in national security issues is reviewed, as well as the domestic and international environment in which involved decisionmakers operated. Finally, the rationale for the test in 1974 is examined. The conclusion revealed is that the explosion of a nuclear device by India in 1974 was primarily related to improving Indian international prestige among Third World countries and uniting a rapidly disintegrating Indian societal consensus. In themselves, individual decision-making theories were found to be of little use, but a combination of the various elements allowed a greater comprehension of the events leading up to the test than might otherwise have been the case.

Decision���Making in Public Good Dilemmas: Theory and Agent���Based Simulation

In many real world contexts individuals find themselves in situations where they have to decide between options of behaviour that serve a collective purpose or behaviours which satisfy one���s private interests, ignoring the collective. In some cases the underlying social dilemma (Dawes, 1980) is solved and we observe collective action (Olson, 1965). In others social mobilisation is unsuccessful. The central topic of social dilemma research is the identification and understanding of mechanisms which yield to the observed cooperation and therefore resolve the social dilemma. It is the purpose of this thesis to contribute this research field for the case of public good dilemmas. To do so, existing work that is relevant to this problem domain is reviewed and a set of mandatory requirements is derived which guide theory and method development of the thesis. In particular, the thesis focusses on dynamic processes of social mobilisation which can foster or inhibit collective action. The basic understanding is that success or failure of the required process of social mobilisation is determined by heterogeneous individual preferences of the members of a providing group, the social structure in which the acting individuals are contained, and the embedding of the individuals in economic, political, biophysical, or other external contexts. To account for these aspects and for the involved dynamics the methodical approach of the thesis is computer simulation, in particular agent-based modelling and simulation of social systems. Particularly conductive are agent models which ground the simulation of human behaviour in suitable psychological theories of action. The thesis develops the action theory HAPPenInGS (Heterogeneous Agents Providing Public Goods) and demonstrates its embedding into different agent-based simulations. The thesis substantiates the particular added value of the methodical approach: Starting out from a theory of individual behaviour, in simulations the emergence of collective patterns of behaviour becomes observable. In addition, the underlying collective dynamics may be scrutinised and assessed by scenario analysis. The results of such experiments reveal insights on processes of social mobilisation which go beyond classical empirical approaches and yield policy recommendations on promising intervention measures in particular.