A boundary tracking method crucial to function approximation-based image coding is proposed that solves the problems caused by duplicate tracking the common boundaries of color regions when methods conventionally appl...
详细信息
A boundary tracking method crucial to function approximation-based image coding is proposed that solves the problems caused by duplicate tracking the common boundaries of color regions when methods conventionally applied to bilevel raster images are inappropriately used. Experimental evaluation is performed to verify its effectiveness.
This work introduces a scheme of layered neural network training, which incorporates a dynamical model alteration during training, and regularization of the features extracted in the hidden layer units. So far, use of...
详细信息
This work introduces a scheme of layered neural network training, which incorporates a dynamical model alteration during training, and regularization of the features extracted in the hidden layer units. So far, use of Model Switching (MS), which is a simultaneous search scheme for an optimal model and parameter, proved to improve training efficiency and generalization ability as a side effect. In MS, the operation to switch the network to a different model involve orthogonalization of the features extracted in the hidden layer. Assuming that the orthogonalization contributes to the observed merits, joint use of MS and orthogonalization of the hidden layer feature by introducing a regularization term in the training, is introduced. The network trained by the proposed training scheme is applied to a pattern recognition problem, and some improvement in training efficiency and generalization ability were observed.
In this paper, we looked into two issues in pattern recognition using neural networks trained by back propagation (BP), namely inefficient learning and insufficient generalization. We observed that these phenomena are...
详细信息
In this paper, we looked into two issues in pattern recognition using neural networks trained by back propagation (BP), namely inefficient learning and insufficient generalization. We observed that these phenomena are partly caused by the way the hidden layer units responds to the inputs. In order to solve the issues, we introduced regularization of the hidden layer unit response which amounts to suppressing the correlation among the response of the hidden layer units, and pruning the unit with the method unit fusion. The results of using the proposed technique were compared with the case of conventional technique in pattern recognition problems. From the results of the experiments, the rate of correct recognition increased when using regularization in the hidden layer unit response is performed, and it turned out that the required number of training epochs also decreases.
To match the different resolution capabilities of devices used in the telemedicine network, we must perform resolution conversion for fundus photograph to the required resolution with the information important to opht...
详细信息
To match the different resolution capabilities of devices used in the telemedicine network, we must perform resolution conversion for fundus photograph to the required resolution with the information important to ophthalmologists' diagnosing well preserved. In this paper, in order to keep important information, such as color and shapes of blood vessels, optic disk and lesion areas on the fundus photograph after resolution conversion, we adopt three kinds of fluency functions to approximate shapes and propose a shape-directed resolution conversion method by using two variables compactly supported fluency sampling functions. The method leads the fundus photograph resolution conversion with features of scalability and good information preservation.
This paper discusses a method to determine the optimal node location before a long distance FWA network (22 GHz band, 6.3 Mbit/s) is set. During our 3 years experiment, we found out that the long distance FWA net with...
详细信息
This paper discusses a method to determine the optimal node location before a long distance FWA network (22 GHz band, 6.3 Mbit/s) is set. During our 3 years experiment, we found out that the long distance FWA net with 2 links (about 60 km) was down for approximately 42 hours per year due to heavy rain fall conditions. This means that the long distance FWA net is practicable to some extent. We discuss a method to estimate the working ratio between two unknown locations using AMeDAS weather data. By using this method, an appropriate node location can be decided before FWA net is deployed. This approach can improve the working ratio in the sense of link-up continuation.
In this paper, a Trademark Application Assistant (TAST) that aims to speed up the process of a successful trademark application is proposed. The core of TAST is a similarity-based image retrieval system that makes use...
详细信息
In this paper, a Trademark Application Assistant (TAST) that aims to speed up the process of a successful trademark application is proposed. The core of TAST is a similarity-based image retrieval system that makes use of relaxation matching on extracted contour segments. A novelty of this research lies in the processing architecture introduced for adapting the highly effective but computation expensive relaxation matching algorithm to real-time image retrieval. Preliminary results illustrating the quality of matching and the scalability of the processing architecture in terms of response time are included. The proposed architecture is not limited to using relaxation matching based servers, but is also applicable to other types of matching server as well.
This paper describes a method of power assist control for lower body based on neuromuscular signal, s-EMG (surface ElectroMyogram/Myoelectricity), and impedance adjustment around knee joint with the assist system, HAL...
详细信息
This paper describes a method of power assist control for lower body based on neuromuscular signal, s-EMG (surface ElectroMyogram/Myoelectricity), and impedance adjustment around knee joint with the assist system, HAL (Hybrid Assistive Leg) -3 we have developed. Virtual Torque calculated by s-EMG enabled the HAL-3 to be operated as to intention of the experimental subject put on HAL, and assist the motion of lower body by predicting the moment around joints. Besides, the operator was able to swing the leg lighter by reducing the inertia and viscous friction around joint of the subject and HAL-3. In order to verify the proposed method, experiments for simple motion was performed with impedance values found by parameter identification with RLS (Recursive Least Square) method. The evaluation of assisted motion was done by Assist Efficiency (AE) calculated from s-EMG in nearly proportion to the operator's muscle force. The results showed the response of operational signal into actuator with impedance adjustment was improved dramatically, and the amplitudes of s-EMG were reduced significantly, then we could confirm the availability of impedance adjustment.
In the delayed match-to-sample task, responses of inferior temporal neurons to adjacent stimuli in the sequence are correlated to each other when the monkey was trained repeatedly with the sequence of visual stimuli, ...
详细信息
In the delayed match-to-sample task, responses of inferior temporal neurons to adjacent stimuli in the sequence are correlated to each other when the monkey was trained repeatedly with the sequence of visual stimuli, although the monkey was not required to associate the stimuli with each other. This correlation, however, is not observed for a monkey with lesions of the rhinal cortex, which is not consistently explained by existing models of such correlated responses. In the present study, we construct a model consisting of two networks corresponding to area TE and the perirhinal cortex, and show that perirhinal plasticity may underlie the mechanism of implicit association learning.
We propose an efficient heuristic algorithm that sets up and releases lightpaths for connection requests dynamically. We partition the routing and wavelength assignment (commonly known as RWA) problem into two subprob...
详细信息
ISBN:
(纸本)0780374002
We propose an efficient heuristic algorithm that sets up and releases lightpaths for connection requests dynamically. We partition the routing and wavelength assignment (commonly known as RWA) problem into two subproblems and solves both of them using a well-known shortest path routing algorithm. For solving the routing subproblem, an auxiliary graph is created whereby the nodes and links in the original network are transformed to the edges and vertices, respectively, and the availability of each wavelength on the input and output links of a node as well as the number of available wavelength converters are taken into account in determining the weights of edges. Furthermore, for solving the wavelength assignment subproblem, an auxiliary graph is also utilized and the cost for wavelength conversion is taken into consideration in the edge weight function. A distinguished feature of our algorithm is that it employs more accurate network information on the availability of both the wavelengths and the wavelength converters than the existing algorithms in deciding the routing and the wavelength assignment. Simulation results show that our algorithm performs much better than previously proposed algorithms with comparable computation time, especially when the number of wavelengths is large while the number of converters at each node is limited.
This paper aims at proposing and comparing two fuzzy models and a statistical model for clustering based on L1-space. Clustering methods in the fuzzy models are the standard fuzzy c-means and an entropy regularization...
详细信息
暂无评论