Informatica

Machine Translation has become an important tool in overcoming the language barrier. The quality of translations depends on the languages and used methods. The research presented in this paper is based on well-known standard methods for Statistical Machine Translation that are advanced by a newly proposed approach for optimizing the weights of translation system components. Better weights of system components improve the translation quality. In most cases, machine translation systems translate to/from English and, in our research, English is paired with a Slavic language, Slovenian. In our experiment, we built two Statistical Machine Translation systems for the Slovenian-English language pair of the Acquis Communautaire corpus. Both systems were optimized using self-adaptive Differential Evolution and compared to the other related optimization methods. The results show improvement in the translation quality, and are comparable to the other related methods.

The Matrix Bandwidth Minimization Problem (MBMP) seeks for a simultaneous reordering of the rows and the columns of a square matrix such that the nonzero entries are collected within a band of small width close to the main diagonal. The MBMP is a NP-complete problem, with applications in many scientific domains, linear systems, artificial intelligence, and real-life situations in industry, logistics, information recovery. The complex problems are hard to solve, that is why any attempt to improve their solutions is beneficent. Genetic algorithms and ant-based systems are Soft Computing methods used in this paper in order to solve some MBMP instances. Our approach is based on a learning agent-based model involving a local search procedure. The algorithm is compared with the classical Cuthill-McKee algorithm, and with a hybrid genetic algorithm, using several instances from Matrix Market collection. Computational experiments confirm a good performance of the proposed algorithms for the considered set of MBMP instances. On Soft Computing basis, we also propose a new theoretical Reinforcement Learning model for solving the MBMP.

This paper presents an improved differential evolution (IDE) method for the solution of large-scale unit commitment (UC) problems. The objective of the proposed scheme is to determine the generation schedule which minimizes the total operating cost over a given time horizon subject to a variety of constraints. Through its use of enhanced acceleration and migration processes, the IDE method limits the population size required in the search procedure and is therefore an ideal candidate for the solution of large-scale combinatorial optimization problems. The effectiveness of the proposed approach is verified by performing a series of simulations based upon the practical Tai-Power System (TPS) and various other power systems presented in the literature. In general, the results show that the IDE scheme outperforms existing deterministic and stochastic optimization methods both in terms of the quality of the solutions obtained and the computational cost. Furthermore, it is found that the magnitude of the cost savings achieved by the IDE scheme compared to that obtained by the other optimization techniques increases as the number of generating units within the power system increases. Therefore, the proposed scheme represents a particularly effective technique for the solution of large-scale UC problems.

Reinforcement learning addresses the question of how an autonomous agent can learn to choose optimal actions to achieve its goals. Efficient exploration is of fundamental importance for autonomous agents that learn to act. Previous approaches to exploration in reinforcement learning usually address exploration in the case when the environment is fully observable. In contrast, we study the case when the environment is only partially observable. We consider different exploration techniques applied to the learning algorithm “Utile Suffix Memory”, and, in addition, discuss an adaptive fringe depth. Experimental results in a partially observable maze show that exploration techniques have serious impact on performance of learning algorithm.