Illya Kokshenev

An efficient multi-objective learning algorithm for RBF neural network

Most of modern multi-objective machine learning methods are based on evolutionary optimization algorithms. They are known to be global convergent, however, usually deliver nondeterministic results. In this work we propose the deterministic global solution to a multi-objective problem of supervised learning with the methodology of nonlinear programming. As the result, the proposed multi-objective algorithm performs a global search of Pareto-optimal hypotheses in the space of RBF networks, determining their weights and basis functions. In combination with the Akaike and Bayesian information criteria, the algorithm demonstrates a high generalization efficiency on several synthetic and real-world benchmark problems.

A multi-objective approach to RBF network learning

The problem of inductive supervised learning is discussed in this paper within the context of multi-objective (MOBJ) optimization. The smoothness-based apparent (effective) complexity measure for RBF networks is considered. For the specific case of RBF network, bounds on the complexity measure are formally described. As the synthetic and real-world data experiments show, the proposed MOBJ learning method is capable of efficient generalization control along with network size reduction.

A Web-based Decision Support Center for Electrical Energy Companies

This study describes a framework for a Web-based decision support center (DSC) to aid various interrelated decisionmaking situations, which emerge from planning and management processes in electrical energy companies. This framework supports small collaborative groups working in an asynchronous way, in an environment where a single decision-maker (DM), who centralizes the responsibility for a final decision, can be aided by a group of experts, who contribute with their opinions to that decision. It is based on the construction and the analysis of (X, R) models, where X is a set of feasible solutions and R a set of fuzzy preference relations. Two preference formats, namely value functions and fuzzy sets, are made available to the input of preference information. Transformation functions adequate for dealing with preference measures on interval scales are utilized to construct fuzzy preference relations. The (X, R) models are analyzed by means of procedures based on the use of the Orlovsky choice function. The availability of different aggregation operators allows a DM to reproduce different attitudes: pessimistic, optimistic, compensatory with adjustment of the tradeoff rates among criteria, as well as lexicographic with prioritization of criteria. When the DM cannot choose a unique attitude to analyze a problem, DSC recommends a generalized solution, which considers all attitudes simultaneously. To demonstrate the applicability of the framework, an expansion planning decision-making problem is considered.

A Multi-objective Learning Algorithm for RBF Neural Network

In this paper, the problem of multi-objective supervised learning is discussed within the non-evolutionary optimization framework. The proposed MOBJ learning algorithm performs the search of Pareto-optimal models determining weights,width, prototype vectors, and the quantity of basis functions of the RBF network. In combination with the Akaike information criterion, the algorithm provides high quality solutions.

A flexible multicriteria decision-making methodology to support the strategic management of Science, Technology and Innovation research funding programs

Abstract Research funding programs are a policy instrument utilized by governments to influence the innovation process. They are usually elaborated, launched and managed by research funding agencies. In order to select the most adequate research projects, agencies often rely on the peer review process. This paper introduces a methodology to support funding decisions based on the peer review process. The methodology involves the use of a multicriteria decision model to support the assessment, evaluation, prioritization and selection of applications, under a multi-step decision-making process, which fits into a strategic management cycle within the agency. The Multiattribute Value Theory, being considered under a Value Focused Thinking approach, provides a basis for the construction of the multicriteria decision model. The good practices in peer review and also a logical framework for program management are considered by the methodology. A pilot study, presented in the paper, involved a retrospective implementation of a peer review process in the context of a program launched by the Ministry for Science, Technology, Innovations and Communications and the National Council of Technological and Scientific Development, in Brazil. The methodology allowed a clear distinction of roles. The agency staff in the role of decision-makers was responsible for making value judgments on behalf of the agency. The experts, in the role of committee members and ad hoc reviewers, contributed with their expertise by providing objective assessments. Such assessments served as a basis for evaluating the applications, characterizing the possible portfolios, and can be considered as data in future program evaluation studies.