Informatica

In this paper, we consider the problem of semi-supervised binary classification by Support Vector Machines (SVM). This problem is explored as an unconstrained and non-smooth optimization task when part of the available data is unlabelled. We apply non-smooth optimization techniques to classification where the objective function considered is non-convex and non-differentiable and so difficult to minimize. We explore and compare the properties of Simulated Annealing and of Simultaneous Perturbation Stochastic Approximation (SPSA) algorithms (SPSA with the Lipschitz Perturbation Operator, SPSA with the Uniform Perturbation Operator, Standard Finite Difference Approximation) for semi-supervised SVM classification. Numerical results are given, obtained by running the proposed methods on several standard test problems drawn from the binary classification literature. The performance of the classifiers were evaluated by analyzing Receiver Operating Characteristics (ROC).

The quadratic assignment problem (QAP) is one of the well‐known combinatorial optimization problems and is known for its various applications. In this paper, we propose a modified simulated annealing algorithm for the QAP – M‐SA‐QAP. The novelty of the proposed algorithm is an advanced formula of calculation of the initial and final temperatures, as well as an original cooling schedule with oscillation, i.e., periodical decreasing and increasing of the temperature. In addition, in order to improve the results obtained, the simulated annealing algorithm is combined with a tabu search approach based algorithm. We tested our algorithm on a number of instances from the library of the QAP instances – QAPLIB. The results obtained from the experiments show that the proposed algorithm appears to be superior to earlier versions of the simulated annealing for the QAP. The power of M‐SA‐QAP is also corroborated by the fact that the new best known solution was found for the one of the largest QAP instances – THO150.

The problem of parameter clustering on the basis of their correlation matrix is considered. The convergence in probability of parameter clustering based on the simulated annealing is investigated theoretically.

In this paper we are concerned with global optimization, which can be defined as the problem of finding points on a bounded subset of R^m, in which some real-valued function f(x) assumes its optimal value. We consider here a global optimization algorithm. We present a stochastic approach, which is based on the simulated annealing algorithm. The optimization function f(x) here is discrete and with noise.

It is well known that, in general, exact algorithms for the Quadratic Assignment Problem (QAP) cannot solve problems of size N>15. Therefore, it is necessary to use heuristic approaches for solving large-scale QAPs. In this paper, we consider a class of heuristic approaches based on local search criteria. In particular, we selected four algorithms; CRAFT, Simulated Annealing, TABU search and the Graph Partitioning (GP) approach and studied their relative performance in terms of the quality of solutions and CPU times. All of these algorithms performed roughly the same, based on the results of two sets of test problems executed on an IBM ES/3090-600S computer.

In this paper we deal with the problem of extremal parameter grouping. The problem formulation, the algorithms of parameter grouping and the fields of implementation are presented. The deterministic algorithms of extremal parameter grouping often find the local maximum of the functional, characterizing the quality of a partition. The problem has been formulated as a problem of combinatorial optimization and attempted to be solved using the simulated annealing strategy. The algorithms, realizing such a strategy and devoted to the solving of the problem concerned, are proposed and investigated.