In the minimization of the number of subtours made by the insertion head of an SMD placement machine a variant of the network flow problem arose. In a network with <span class="tex">n</span> vertices and <span class="tex">m</span> arcs a set <span class="tex">F</span> of arcs (parametrized arcs) is given. The task is to find a flow of a given size such that the maximum of flow values along the arcs from <span class="tex">F</span> is minimized. This problem can be solved by a sequence of maximum flow computations in modified networks where the capacities of the parametrized arcs are successively set to an increasing sequence of target parameter values. We show that it suffices to consider at most <span class="tex">O(|F|)</span> different target values and so this approach leads to a strongly polynomial algorithm consisting of at most <span class="tex">O(|F|)</span> maximum flow computations.
A well known problem in unit selection speech synthesis is designing the join and target function sub-costs and optimizing their corresponding weights so that they reflect the human listeners' preferences. To achieve this we propose a procedure where an objective criterion for optimal speech unit selection is used. The objective criterion for tuning the cost function weights is based on automatic speech recognition results. In order to demonstrate the effectiveness of the proposed method listening tests with 31 naive listeners were performed. The experimental results have shown that the proposed method improves speech quality and intelligibility. In order to evaluate the quality of synthesized speech the unit selection speech synthesis system is compared with two other Croatian speech synthesis systems with voices built using the same recorded speech corpus. One of these voices was built with the Festival speech synthesis system using the statistical parametric method and the other is a diphone concatenation based text-to-speech system. The comparison is based on subjective tests using MOS (mean opinion score) evaluation. The system using the proposed method used for cost function weights optimization performs better than other compared systems according to the subjective tests.
The predictive performance of Echo State neural networks were optimized for electrical load forecasting and compared to the results achieved by competitors in the worldwide Eunite Competition #1. The test data used were the actual results of the competition, attached to a specific region. A regular adaptation of an \textit{Echo State }neural network was optimized by adapting the weights of the dynamic reservoir through Anti-Hebbian learning, and the weights from input and output neurons to the hidden neurons were optimized using the Metropolis algorithm. The results achieved with such an optimized Echo State neural network would gain a strong second place within the Eunite competition.
Combinatorial optimization problems are extensively solved by using neural networks. Hopfield-Tank model is used to solve Traveling Salesman Problem and many NP-Hard Problems. This paper describes a neural network optimizer/scheduler that optimizes a solution for a highly complicated version of N Queens Problem (NQP), i.e. N+1 non-threatening Queens on a N*N chessboard with an intermediate pawn on it. Both synchronous and asynchronous methods of updating of the neurons have been applied for optimization of N+1 Queens Problem. Computer simulations are used to confirm the results. The proposed neural network is attracted to optimized solution or finds the global minima in 90% of the trials. A new rule of initialization, i.e. the proximity rule of initialization has been proposed. Using the proximity rule of initialization the performance of the system is enhanced and the system converges to an optimal solution in much less time. Many novel applications like multiprocessor job scheduling, resource optimization, of the above mentioned algorithm have been proposed. N Queens Problem has been solved by many techniques but no other algorithm exists to solve N+1 QP in the literature. Consequently, the performance of the network is compared with full space search algorithm.
This paper deals with issues related to convenient monitoring of subsidence due to longtime mining activities of Czech Karviná Mine: Lazy plant, using satellite SAR interferometry (InSAR) techniques. It maintains approaches for optimizations of differential InSAR, especially including available filtering possibilities. It was realized that current SAR satellites were not able to appropriately evaluate deep subsidence as it occurs in Czech mining sites. Other issues are related to the presence of dense vegetation. Data from previous and only available L-band SAR satellite ALOS demonstrates its high potential in this area of interest. However, only a few acquisitions are available of the mining site disallowing usage for purposes of continuous monitoring of subsidence in the area. Processing results of InSAR techniques of the Lazy plant are presented in this paper., Milan Lazecký and Eva Jiránková., and Obsahuje bibliografii
Fuzzy min-max neural network (FMN), proposed by Simpson is a well-known supervised neuro-fuzzy classifier that has been successfully used by many researchers for pattern recognition. However, the FMN represents the learned knowledge with exhaustive details in a `fine-grained' manner that reduces its performance for pattern recognition in terms of the recall time per pattern. In this paper, we adapt the basic architecture of the FMN to represent the learned knowledge in a compact way that is in a `coarse-grained' manner, which is closed to human thinking. The working of the proposed method that is fuzzy min-max neural network with knowledge compaction (FMN-KC) is illustrated using the Fisher Iris dataset. The potential of using the FMN-KC for supervised outlier detection is demonstrated using a time-series disk defect dataset published by NASA and KDD cup 99 dataset available in UCI repository. The proposed method achieves around 50% gain in the recall time as compared to the original FMN and the recognition rate is also comparable. We strongly recommend using the proposed architecture FMN-KC for supervised outlier detection in the real time applications, where recall time per pattern is one of the key parameters.
This paper is concerned with the design of event-based state estimation algorithm for nonlinear complex networks with fading measurements and stochastic coupling strength. The event-based communication protocol is employed to save energy and enhance the network transmission efficiency, where the changeable event-triggered threshold is adopted to adjust the data transmission frequency. The phenomenon of fading measurements is described by a series of random variables obeying certain probability distribution. The aim of the paper is to propose a new recursive event-based state estimation strategy such that, for the admissible linearization error, fading measurements and stochastic coupling strength, a minimum upper bound of estimation error covariance is given by designing the estimator gain. Furthermore, the monotonicity relationship between the trace of the upper bound of estimation error covariance and the fading probability is pointed out from the theoretical aspect. Finally, a simulation example is used to show the effectiveness of developed state estimation algorithm.
Trust-aware recommender system (TARS) recommends ratings based on user trust. It greatly improves the conventional collaborative filtering by providing reliable recommendations when dealing with the data sparseness problem. One basic research issue of TARS is to improve the recommending efficiency, in which the key point is to find sufficient number of recommenders efficiently for active users. Existing works searched recommenders via a skeleton, which consists of a number of hub nodes. The hub nodes are those who have superior degrees based on the scale-freeness of the trust network. However, existing works did not consider the skeleton maintenance cost and the coverage overlap between nodes of the skeleton. They also failed to suggest the proper size of the skeleton. This paper proposes an optimized TARS model to solve the problems of existing works. By using the genetic algorithm, our model chooses the most suitable nodes for the skeleton of recommender searching. It can achieve the maximum prediction coverage with the minimum skeleton maintenance cost. Simulation results show that compared with existing works, our model can reduce more than 90{\%} of the skeleton maintenance cost with reasonable prediction coverage.
The purpose of the paper is to discuss the applicability of stochastic programming models and methods to civil engineering design problems. In cooperation with experts in civil engineering, the problem concerning an optimal design of beam dimensions has been chosen. The corresponding mathematical model involves an ODE-type constraint, uncertain parameter related to the material characteristics and multiple criteria. As a result, a multi-criteria stochastic nonlinear optimization model is obtained. It has been shown that two-stage stochastic programming offers a promising approach to solving similar problems. A computational scheme for this type of problems is proposed, including discretization methods for random elements and ODE constraint. An approximation is derived to implement the mathematical model and solve it in GAMS. The solution quality is determined by an interval estimate of the optimality gap computed by a Monte Carlo bounding technique. The parametric analysis of a multi-criteria model results in efficient frontier computation. Furthermore, a progressive hedging algorithm is implemented and tested for the selected problem in view of the future possibilities of parallel computing of large engineering problems. Finally, two discretization methods are compared by using GAMS and ANSYS.