The purpose of our work is to design a financial analysis system by an intuitive reasoning on an optimistic-pessimistic axis. The intuitive reasoning is totally different from a logical reasoning used in AI and is defined by subjective information such as adjectives high, low, good, bad and etc. A Computer-Aided Financial Analysis Systems (CAFAS) based on fuzzy theory and the intuitive reasonings are a new challenge to computer technology experts and management science specialists. In this paper, we propose a new financial analysis system based on the intuitive reasonings. Firstly, we describe the intuitive reasonings. Secondly, the intuitive reasoning is applied to financial data. Thirdly, the results of the intuitive reasoning are projected on the optimistic-pessimistic axis. In the projection, fuzzy membership functions for financial analysis are calculated. Finally, practical analyses of four computer enterprises are given in this paper and four membership functions of typical financial condition for financial analysts are shown.
In this paper, we propose the dual-width windowed segment (DWWS) which consists of a short cepstrum segment and a long Δ cepstrum segment, as a more effective segmental input vector than the conventional one. First, using a criterion for cluster analysis, we evaluate various compositions of feature vectors based on their ability of phoneme separation to show the effectiveness of DWWS. Then we carry out discrete HMM speech recognition experiments to verify the result of evaluation. As a result, it is shown that the DWWS brings on high recognition performance when a categories-dependent codebook is used for vector quantization.
Recently, Just-In-Time (JIT) production systems have much influence on the production fields. Miyazaki et al. propose a concept of actual flow time, which is a performance measure for scheduling in a JIT production environment. Scheduling problems are one of representative combinatorial optimization problems. Hopfield and Tank show that some combinatorial optimization problems can be solved by the artificial neural network system. Arizono et al. propose a neural solution for minimizing total actual flow time by the Gaussian machine. However, their method retains some problems which originate in the analog neurons. Then, we use interconnected neural networks which consist of the binary neurons whose output states take values either 0 or 1 unlike the Arizono's system. We call such a network the binary neural network. This paper deals with the scheduling problem for minimizing total actual flow time by the binary neural network.
It is known that ants and bees are working cooperatively as a group by detecting the information transmission material called pheromone. In this paper, this idea is combined with meta-heuristics search, and applied to the problem of finding an optimal operational sequence. In modeling, one job is treated as an agent such as ant or bee. Each agent scatters its peculiar pheromone on each processing order corresponding to decision variable, and exchanges the orders statistically among agents according to the distribution of pheromone. The results of the exchanges are evaluated as a whole and the update value of each pheromone is determined. We propose a method that generates a processing sequence converging to the optimal processing order by regulating the related parameters through such iteration. We have developed a computer program based on the algorithm. The usefulness of the proposed method is confirmed by numerical experiments for scheduling of the hot rolling process.
In our previous works, we have proposed “virtual passive dynamic walking” with virtual gravity for biped robots in order to realize active walking on the level ground without any gait design in advance. In this paper we discuss some control problems of a kneed biped robot and propose “modified compass-like virtual passive dynamic walking” with active knee-lock algorithms in order to avoid “foot scuffing” problem during the single support phase. Furthermore, a virtual coupling control law is proposed which can realize variable walking pattern with respect to the robot's energy levels. By the effect of the control law, the robot which is a hybrid dynamical system can be regarded as a passive system which does not include any collisions, and variable walking pattern can be realized without loss of properties of a virtual passive walk. The validity of the proposed methods has been examined by numerical simulations.
This paper considers an off-line reference management technique for constraint fulfillment. The proposed finite dimensional optimization program provides the managed reference signal which assures infinite-time constraint fulfillment. The main idea is to restrict the state variable at the final instance of finite-time reference management to a certain invariant subset of the state space such that fulfilling constraints for the primary designed closed-loop system is equivalent to restricting the state dynamics to this set. The proposed technique also, enables us to improve the control performance of systems with constraints. The control scheme is proved to fulfill the specified state and control constraints and shown to improve the set-point tracking performance.