|
1. |
A speculative execution scheme of macrotasks for parallel processing systems |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 1-15
Hayato Yamana,
Toshiaki Yasue,
Yoshihiko Ishii,
Yoichi Muraoka,
Preview
|
PDF (1245KB)
|
|
摘要:
AbstractThis paper considers the high‐speed execution of FORTRAN programs on parallel processing systems and proposes the parallelizing scheme of the program and execution based on the speculative execution over multiple conditional branches. Several techniques have been proposed that parallelize the program including conditional branches.A method which does not use the speculative execution is: (1) the method called earliest execution condition determination. As the methods which use the speculative execution are: (2) speculative evaluation scheme for a single conditional branch for the superscalar processor or VLIW computer; and (3) multiple speculative execution scheme assuming particular loops.There are the following problems: (1) sufficient parallelism is not extracted only by determining the earliest execution condition; (2) the speed improvement that can be realized by the speculative execution of a single conditional branch is at most twofold; and (3) the scheme can be applied only to particular loops.This paper divides the program into macrotasks, and defines the multiple stage speculative execution scheme between macrotasks on the general parallel processing system. Then, the macrotask execution control for the individual macrotask is proposed, using the execution start condition, the control establishment condition and the execution stop conditio
ISSN:0882-1666
DOI:10.1002/scj.4690260601
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
2. |
Design and implementation of load‐balancing mechanism at command level on UNIX |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 16-26
Nariyoshi Yamai,
Susumu Wakabayashi,
Shinji Shimojo,
Hideo Miyahara,
Preview
|
PDF (1138KB)
|
|
摘要:
AbstractThis paper proposes a design and implementation of a new load‐balancing mechanism at the command level on UNIX operating systems. In this mechanism, only the command interpreter bash is modified to have a new command executed on the lightest‐loaded computer, and efficient load balancing is made without modifying the UNIX kernel or application program.In the author's implementation, load is measured by statistics information server rstatd on each computer when a new command arrives, and the new command is assigned to the lightest‐loaded computer among the local host and the computers that respond faster than it. Using this method, loads can be balanced even when there is no idle computer, and the system can be made fault‐tolerant in the case where some machines are down. This mechanism also has network transparency, such as file location transparency and preservation of command execution environment, by using REX for remote execution, which is provided on many UNIX systems as one of the RPC (remote procedure call) s
ISSN:0882-1666
DOI:10.1002/scj.4690260602
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
3. |
Human‐face search and location in a scene by multi‐pyramid architecture for personal identification |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 27-38
Makoto Kosugi,
Preview
|
PDF (1888KB)
|
|
摘要:
AbstractTo realize fully automated face image recognition, there must be thorough processing from the detection of the face in a scene to recognition. It is very difficult, however, to apply the detection or recognition process based on the shape features as in the conventional method to the face which is a typical soft object.This paper aims at personal identification by the facial image. The face in a scene is sought by coarse‐to‐fine processing using only the gray‐level data, and the result is applied to the recognition.First, the human head is detected from the scene using the multiresolution mosaic. Then the central part of the face is detected from the head region using the mosaic, and the precise position is determined based on the histogram for the eye and the nose region. The search algorithm is applied to 100 personal images derived from the motion image. The detection and location succeeded in 97 percent of the cases, except for the face with eyes shielded by hair, for example. When the result of successful detection is applied to the recognition, the recognition rate of 99 percent is obtained. In this method, a facial image of any size at any position in a scene can be detected. Other features are that the background can be uniform, and the color data are not required, which greatly relaxes the past requirement for the input
ISSN:0882-1666
DOI:10.1002/scj.4690260603
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
4. |
Optical model for motion image analysis |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 39-48
Naoki Mukawa,
Preview
|
PDF (1149KB)
|
|
摘要:
AbstractIn previous analysis of the motion image, the existence of the light source has not particularly been considered. In this paper, the optical model describing the relations among the light source, the object and the view point is introduced. The optical flow estimation problem is discussed based on the gradient method, using the constructed model. First, the relation between the light source and the object model is described as an optical model, and the constraint inherent in the considered scene is derived. As a result, it is shown that the optical flow can be estimated with a high precision also for the scene containing a light source.It is shown that the parameter, which has been considered as a correction term for improving the precision of the optical flow, is a physical parameter indicating the geometrical relation between the light source and the object, from the standpoint of the optical model. Then, not only can the optical flow be estimated with a high precision, but also the information concerning the incident direction from the light source and the reflection coefficients of the object can be newly restored. If the scene is of a simple structure, all parameters that describe the scene, i.e., the information concerning the location of the light source, the object and the camera, as well as the information concerning the shape of the object, and the reflection coefficient of the surface of the object can be estimated. The feasibility of the scene reconstruction is investigated using the CG techniques.
ISSN:0882-1666
DOI:10.1002/scj.4690260604
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
5. |
Analysis of the error back‐propagation learning algorithms with gain |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 49-58
Qi Jia,
Katsuyuki Hagiwara,
Shiro Usui,
Naohiro Toda,
Preview
|
PDF (768KB)
|
|
摘要:
AbstractAs the method to accelerate the learning by error back‐propagation, several studies have been proposed in which the parameter called gain is introduced. In those studies, however, the acceleration effect is evaluated only numerically, and there is no theoretical analysis of the effect of the gain on the learning process.This paper points out that those studies can also be realized by methods without introducing the gain, and presents a detailed analysis of the effect of the gain from a unified viewpoint. The following properties are revealed as a result. The error back‐propagation method, in which a constant gain is introduced, can be reduced to the ordinary error back‐propagation method without introducing the gain. When the dynamic gain is introduced, the method cannot be reduced to the steepest descent method, as well as the momentum method, without introducing the gain. Furthermore, it is shown that there exists a characteristic superellipse that determines the behavior of the gain.By analyzing the characteristic superellipse, a theoretical basis is provided for the instability of the method introducing the dynamic gain. This paper presents a unified treatment of the method introducing the gain and the method not introducing the gain from a unified viewpoint which have been considered independently. The effect of the gain on the learning process is analyzed, which will help in developing a new learning method in the f
ISSN:0882-1666
DOI:10.1002/scj.4690260605
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
6. |
Automatic tracking of highway road edge based on vehicle dynamics |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 59-70
Shinji Negishi,
Shinji Ozawa,
Masataka Chiba,
Preview
|
PDF (1099KB)
|
|
摘要:
AbstractThere are many efforts being made toward the visual navigation system for the autonomous vehicle. When the highway road is considered, where the constraint for the road structure can be utilized, it is effective to extract the white line (lane marker), which is enforced to be drawn as the borderline of the lane.This paper considers the automatic extraction of the road edge from the white line in the road image taken by the vehicle camera. There is already proposed a technique which assumes that the road edge image is extracted from the white line, and estimates the horizontal curvature of the road and the vibration component of the vehicle (dynamic angle of the vehicle). It is adequate to use the break (end point) of the white dashed line as the feature point for estimation of the vehicle angle. To extract the road edge from the dashed line, however, it is required to trace the dashed line which moves between consecutive frames.From such a viewpoint, this paper utilizes the result obtained by the already proposed estimation method for the vehicle angle, and predicts the motion of the end‐point of the dashed line on the image due to the motion of a vehicle. By restricting the range of search based on the result, the end‐point is tracked. An experiment was conducted for the sequence of the image taken for the actual road. It is verified that a sufficient accuracy is obtained for the position prediction, so that the restricted range of search contains the end‐point, which will help the automatic extraction of the road edge. Then, a system is constructed as a cooperative system, which estimates the vehicle angle the road shape from the real image, using the automatically extracted road edge.It is verified by an on‐line experiment for the real image that the vehicle angle and the horizontal curvature of the road can effectively be estimated. Those results are reported in thi
ISSN:0882-1666
DOI:10.1002/scj.4690260606
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
7. |
Construction of structural edge map on chest radiograph using hough transformation and line connection |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 71-78
Tetsuo Sugahara,
Yoshio Yanagihara,
Naozo Sugimoto,
Preview
|
PDF (1317KB)
|
|
摘要:
AbstractFor computer diagnosis of a chest radiograph, we propose a method to make a structural edge map consisting of subclaviculars, heart, diagram, and ribs that have been extracted from a chest radiograph. On processing, the intensity of the edges of an input image was calculated by a derivative filter. Next, the possibility of lines and/or arcs was calculated by Hough transformation. This transformation was applied to 13 small areas for each of the organs in the image. A reconnection of lines and arcs representing the edges of the organs was performed, and a structural edge map was then constructed. As a result detection of subclaviculars, heart, diaphragms, and inferior edge lines of ribs was very good (ge; 79 percent). However, detection on superior edge lines was poor (= 55 percent).
ISSN:0882-1666
DOI:10.1002/scj.4690260607
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
8. |
Networked bubble propagation method as a polynomial‐time hypothetical reasoning for computing near‐optimal solution |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 79-94
Yukio Ohsawa,
Mitsuru Ishizuka,
Preview
|
PDF (1229KB)
|
|
摘要:
AbstractHypothetical reasoning is important to knowledge system framework due to both its theoretical basis and its usefulness in solving practical problems including diagnosis, design, etc. One crucial problem with hypothetical reasoning is, however, its slow inference speed. To achieve practical or tractable speed, an approximate solution method of 0–1 integer programming named Pivot and Complement has been applied so far to a cost‐based hypothetical reasoning, in which a numerical weight is assigned to each possible element hypothesis and an optimal solution hypothesis‐set with the minimal sum of its element hypotheses' weights is searched. In this method, all described knowledge is regarded as constraints. Then, the restricted part of knowledge is transformed into inequalities to apply 0–1 integer programming. While the computational complexity of hypothetical reasoning is NP‐complete or NP‐hard, an approximate solution method of 0–1 integer programming allows polynomial inference time in the order ofN4for finding a near‐optimal solution hypothesis, whereNis the number of possible element hypotheses.In this paper, in order to understand the mechanism of the foregoing method from the viewpoint of knowledge processing, it is reformalized using a new type of knowledge network. Our new networked method, name Networked Bubble Propagation is shown to have a similar (or superior) performance to the aforementioned one, i.e., the solutions it obtains are as good as those found with the Pivot and Complement method and the inference time is in the order ofN2. This speed‐up is an advantage of considering the knowledge structure of a given problem, than
ISSN:0882-1666
DOI:10.1002/scj.4690260608
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
9. |
Performance characteristics of analog metal‐oxide semiconductor type neuroprocessor |
|
Systems and Computers in Japan,
Volume 26,
Issue 6,
1995,
Page 95-106
Yoshihide Inoue,
Hideo Taguchi,
Eizo Kuroda,
Preview
|
PDF (804KB)
|
|
摘要:
AbstractTechnological requirements for excellent parallel distributed processing capability of central nervous system has increased recently. This has led to the development of neurosimulators, e.g., neurocomputers and neurochips integrated from artificial neuron combinations. However, in developing the neurosimulators, there is no established theory as to how to establish each individual artificial neuron activity function. In the development of a theory, artificial neuron combinations were trial‐produced using analog MOS devices. The characteristics of the Analog MOS artificial neurons are as follows:1They consist of equivalent electronic circuits for synapses on the dendrites, cell bodies, and axons, assuming the electrical properties of each section of the bioneuron produces a unique processing pattern.2They exhibit the same electrical performance (absolute refractory period, temporal and spatial weightings of the post synaptic potential, and pulse generation at the axon base) as the bioneurons.3The synaptic equivalent circuits are connected with learning systems to enable plastic input‐output characteristics by changing pulse transmission efficiency according to the input pulse frequencies. In the final research stage, the possibility of fundamental operations, such as filtering and phase shifting in the pulse frequency domain using a neuroprocessor comprising several analog MOS artificial neurons with different pulse transmission efficiencies was investiga
ISSN:0882-1666
DOI:10.1002/scj.4690260609
出版商:Wiley Subscription Services, Inc., A Wiley Company
年代:1995
数据来源: WILEY
|
|