In this paper a new method to detect road signs from color road images and to discriminate them is proposed. This problem has been recognized as an important research field of ITS(Intelligent Transportation Systems). We use two neural networks called Color NN and Shape NN in order to increase the success rate in discrimination. Color NN is used to extract pixels which have colors on road signs. Shape NN, whose inputs are shape feature coefficients, examines whether the regions extracted by Color NN have similar shapes to road signs, i.e., circle, rectangular, and so on. Experimental results show this method is sufficiently efficient for practical use.
This paper describes the generation of story texts from some pictures and transformation of a theme music based on impressions of the generated story. Story texts are generated from information on some pictures. The generation of stories considers the contents of pictures and the connective consistent relationship between pictures. The case based reasoning method is applied to the generation of story texts. The transformed theme music is used as a back ground music fitting the generated story. Neural network models are applied to transform the theme music in order to reflect user's feelings for music and stories. This paper also describes the evaluation experiments to confirm whether the transformed theme music reflects impressions of generated story scenes appropriately or not.
The dynamic programming (DP) is one of the most famous solution methods for a knapsack problem. It is known that DP can obtain the optimal solution at the pseudo-polynomial time, and can solve almost all existing instances. Therefore, the 0-1 knapsack problem is believed to be one of the “easier” NP-hard problems. The currently most successful algorithm for a knapsack problem was presented by Martello et al. based on DP. This algorithm can be seen as a combination of many different concepts and is hence called Combo. Pisinger which is one of the proposers of Combo showed that the knapsack problem still was hard to solve for this algorithm for a variety of new test problems. In this paper, we propose the new improved causal level that flexibly corresponds to various different type problems, and we propose an improved adjustment type genetic algorithm that aims at improving the solution accuracy and shortening the calculation time. The algorithm can adjust a search area in consideration of the stability of each item which can obtain from the greedy algorithm. We apply the proposed method to various difficult instances, and test the effectiveness.
There have been proposed several methods to derive a satisficing solution of the decision maker for multiobjective stochastic programming problems based on the expectation optimization model and the variance minimization model. Satisficing solutions based on these models, however, do not always satisfy the expected utility maximization principle such that the decision maker aims to maximize the expectation of the utility function to express the satisfaction degree of the decision maker for objective functions in uncertain decision making situations. In the meanwhile, there exists a concept to make an ordering of random variables using the second-order distribution function which is the integral of the distibution function, called the second-order stocastic dominance. When the utility function of the decision maker is risk-averse, the second-order stocastic dominance is consistent with the expected utility maximization principle. Therefore, in this paper, we focus on multiobjective stochastic programming problems. After the definition of Pareto optimality based on the second-order stocastic dominance, we propose an interactive fuzzy satisficing method to derive satisficing solutions which are consistent with the expected utility maximization principle for multiobjective stochastic programming problems.
In this paper, we propose an IF... THEN... rule-based inference method which is necessary to construct a natural language communication system and an expert system, and so on. The method is referred to how to estimate a truth qualifier τB' when an input proposition is “QA are F is τ” and IF... THEN... rule “IF Q'A' are F' is τA THEN Q''A'' are F'' is τB” is given and an inference result is “Q''A'' are F'' is τB' ” (Q, Q', Q'': Fuzzy quantifiers, A, A', A'': Fuzzy subjects, F, F', F'': Fuzzy predicates, τ, τA, τB, τB': Truth qualifiers). We propose a method which infers a result proposition “Q''A'' are F'' is τB' ” for monotone Q's and show concrete application examples of the method. Furthermore we compare the inference results under various implication functions used for getting a truth value fuzzy set of the rule.