| Co-Investigator(Kenkyū-buntansha) |
SUGIMOTO Osamu Department of Electrical Engineering, Stanford University, Visitng Scholar, 電気工学科, 客員研究員
KOBAYASHI Hiroaki Graduate School of Information Sciences, Tohoku University, Associate Professor, 大学院・情報科学研究科, 助教授 (40205480)
HAGIWARA Masafumi Department of Electrical Engineering, Keio University, Assistant Professor, 理工学部, 専任講師 (80198655)
GOTOH Eisuke National Institute of Health, Research Associate, 遺伝子資源室, 厚生技官 (90215484)
FUKASE Masa-aki Research Institute of Electrical Communication, Tohoku University, Associate Pro, 電気通信研究所, 助教授 (10125643)
HASEGAWA Katsuo Computer Center, Tohoku University, Associate Professor, 大型計算機センター, 助教授 (70004463)
FLYNN Michael j. E.E.Department, Stanford University, Professor, 計算機システム研究所, 教授
|
|---|
| Research Abstract |
We have planned the joint research project between Tohoku University and Stanford University with the aim of analyzing, synthesizing, and evaluating the performance of brain-structured supercomputers. During the fourth academic year of the project, ten meetings have been held. The meetings have been very profitable for the progress of our project. We have published many papers about brain-structured supercomputers in this year. Following are main subjects discussed extensively and deeply in these meetings.
1. Synthesis of brain-structured supercomputers : The model of a brain-structured supercomputer has been diagramed by using mind computer, expression recognition associative memory, wave cybernetics, artificial cochlea, sparse adaptive memory, wave pipeline, jet pipeline, logical architecture, symbolic architecture, functional architecture, distributed shared memory multiprocessor, and computer graphics etc.
2. Architecture of the sparse adaptive memory : The wave pipcline has been designed by using a CMOS VLSI vector unit. Advances and problems in high speed processor design has been made clear.
3. Analysis of RIGHT computer : So far developed methodologies have been farther investigated to implement possible brain functionalities onto supercomputers from various aspects like memory system, conversion system from input to memory reference frame, mechanisms of heuristics, etc.
4. Architecture of superparallel symbol processing : A book has been published that describes the basis of the VLSI architecture for the superparallel symbol processing.
5. Performance evaluation of LEFT and RIGHT computer : The whole system of computer graphics has been investigated by using multiprocessing techniques for ray tracing and multipass rendering.
6. Study of neural network toward RIGHT computer : Four papers have been published about combination of neural networks and fuzzy inference, knowledge processing by distributed representation.
|
