Papers - YAMAMORI Kunihito
-
超並列計算機上でのニューラルネットワークの並列学習法 Reviewed
山森一人，阿部亨，堀口進
電子情報通信学会論文誌 J80-D-II ( 1 ) 350 - 353 1997.12
Language：Japanese Publishing type：Research paper (scientific journal)
-
準概略図を用いたハイブリッドニューラルネットワークによる3次元物体認識 Reviewed
猪瀬弘和，山森一人，阿部亨，堀口進
電子情報通信学会論文誌 J78-D-II ( 12 ) 1927 - 1931 1995.12
Language：Japanese Publishing type：Research paper (scientific journal)
-
Efficient design of fault-tolerant artificial neural networks Reviewed
Yamamori K., Horiguchi S., Kim J., Park S., Ham B.
IEEE International Conference on Neural Networks - Conference Proceedings 3 1487 - 1491 1995.12
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：IEEE International Conference on Neural Networks - Conference Proceedings
This paper focuses on the study of efficient fault-tolerant design methods in an artificial neural network (ANN) implemented on a digital VLSI chip. Due to the high fault-tolerant capability of biological neural networks, it is widely taken for granted that ANN should also be fault-tolerant. However, if a faulty neuron or a faulty link occurs in ANN currently used in engineering fields, typically ANN no longer carries out the specified performance. The ability of ANN to achieve fault-tolerance, is not inherent, but must be built in. Also, the built-in fault-tolerant mechanism must be practical and efficient enough for a VLSI chip implementation. In this paper, the Partial Retraining scheme is proposed as a design method to achieve fault-tolerance in ANN. The Partial Retraining scheme is applied to only each single neuron which is affected by the hardware fault, not an entire multi-layer network. Therefore, the convergence speed of the Partial Retraining will be much faster than that of the normal learning of the entire multi-layer network. Furthermore, the Partial Retraining can be executed parallelly. To evaluate the performance of the Partial Retraining scheme, we applied the scheme to a large scale ANN for face image recognition.
-
EWS‐FLI‐1 and EWS‐ERG chimeric mRNAs in Ewing's sarcoma and primitive neuroectodermal tumor
Ida K., Kobayashi S., Taki T., Hanada R., Bessho F., Yamamori S., Sugimoto T., Ohki M., Hayashi Y.
International Journal of Cancer 63 ( 4 ) 500 - 504 1995.11
Language：Japanese Publishing type：Research paper (scientific journal) Publisher：International Journal of Cancer
Thde t( 11;22)(q24;q 12) and t(21;22)(q22;q 12) are specific chromosomal translocations found in the Ewinhd family of tumors including ES, PNET and Askin tumors. In these translocations, the amino‐terminal portion of the EWS gene located in 22q 12 fuses to the carboxyl‐terminal portion of the FLI‐I gene located in 11 q24 or the ERG gene located in 21 q22, which belong to the ets oncogene super‐family of transcription activators. We investigated the chimeric mRNAs of 15 ESs (7 cell lines and 8 tumor samples) and 7 PNETs (3 cell lines and 4 tumor samples) using the RT‐PCR method and sequencing. We detected 2 types of EWS‐ERG chimeric mRNA in 2 ES cell lines and I PNET tumor sample in addition to 4 types of EWS‐FU‐1 chimeric mRNA in 11 ESs (4 cell lines and 7 tumor samples) and 4 PNETs (2 cell lines and 2 tumor samples). There seemed to be no association between the type of chimeric mRNA and clinical features such as sex, age, primary site and histopathology of the patients. All of the chimeric mRNAs are generated from in‐frame junctions and are thought to encode fusion proteins that may be the molecular mechanism involved in the Ewing family of tumors. Copyright © 1995 Wiley‐Liss, Inc., A Wiley Company
-
The Efficient Design of Fault-Tolerant Artificial Neural Networks Reviewed
K.Yamamori, S.Horiguchi, J.H.Kim, Sung-K Park, Byung H.Ham
Proc. IEEE International Conference on Neural Networks 1487 - 1491 1995.11
Language：English Publishing type：Research paper (international conference proceedings)
-
The Fault-Tolerant Design of Artificial Neural Networks Reviewed
J.H.Kim, K.Yamamori, S.Horiguchi
Proc. Joint Technical Conference on Circuit/Systems, Computers and Communications 687 - 690 1995.7
Language：English Publishing type：Research paper (international conference proceedings)