A Learning Algorithm with Activation Function Manipulation for Fault Tolerant Neural Networks

スポンサーリンク

概要

• 論文の詳細を見る

• In this paper we propose a learning algorithm to enhance the fault tolerance of feedforward neural networks (NNs for short) by manipulating the gradient of sigmoid activation function of the neuron. We assume stuck-at-0 and stuck-at-1 faults of the connection link. For the output layer, we employ the function with the relatively gentle gradient to enhance its fault tolerance. For enhancing the fault tolerance of hidden layer, we steepen the gradient of function after convergence. The experimental results for a character recognition problem show that our NN is superior in fault tolerance, learning cycles and learning time to other NNs trained with the algorithms employing fault injection, forcible weight limit and the calculation of relevance of each weight to the output error. Besides the gradient manipulation incorporated in our algorithm never spoils the generalization ability.

• 社団法人電子情報通信学会の論文
• 2001-07-01

著者

• HATA Yutaka

  Graduate School of Engineering, University of Hyogo

• Hata Y

  Graduate School Of Engineering University Of Hyogo

• Hata Yutaka

  The Authors Are With Department Of Computer Engineering Himeji Institute Of Technology : The Author

• KAMIURA Naotake

  The authors are with the Department of Computer Engineering, Himeji Institute of Technology

• TANIGUCHI Yasuyuki

  The authors are with the Department of Computer Engineering, Himeji Institute of Technology

• MATSUI Nobuyuki

  The authors are with the Department of Computer Engineering, Himeji Institute of Technology

• Kamiura N

  Graduate School Of Engineering University Of Hyogo

• Kamiura Naotake

  The Authors Are With Department Of Computer Engineering Himeji Institute Of Technology

• Matsui N

  Graduate School Of Engineering University Of Hyogo

• Taniguchi Yasuyuki

  The Authors Are With The Department Of Computer Engineering Himeji Institute Of Technology

• Hata Y

  Himeji Institute Of Technology Japan

関連論文

• Cerebral cortex segmentation with adaptive fuzzy spatial modeling in 3.0T IR-FSPGR MR images (特集 医用システム)
• An Ultrasonic and Air Pressure Sensing System for Detection of Behavior before Getting out of Bed Aided by Fuzzy Theory
• Fuzzy ultrasonic imaging system for visualizing brain surface and skull considering refraction (特集 医用システム)
• Ultrasonography System Aided by Fuzzy Logic for Identifying Implant Position in Bone(Computation and Computational Models)
• A New Ultrasonic Oscillosensor and Its Application in Biological Information Measurement System Aided by Fuzzy Theory(Biological Engineering)
• Computer-Aided Diagnosis of Intracranial Aneurysms in MRA Images with Case-Based Reasoning(Biological Engineering)
• Unconstrained Evaluation System for Heart Rate Using Ultrasonic Vibrograph
• Automated Extraction System of Embedded Tubes from Pulse Radar Image Based on Fuzzy Expert System(Systems and Control)
• D-16-10 An Improved Small Region Growing Method for Segmenting 3-D Liver region from Multidetector CT Images
• Ultrasonic Nondestructive Evaluation for Embedded Objects in Concrete Aided by Fuzzy Logic(Regular Section)
• A Learning Algorithm with Activation Function Manipulation for Fault Tolerant Neural Networks
• On a Weight Limit Approach for Enhancing Fault Tolerance of Feedforward Neural Networks
• Design of Multiple-Valued Programmable Logic Array with Unary Function Generators
• Design and Fault Masking of Two-Level Cellular Arrays on Multiple-Valued Logic
• On Ternary Cellular Arrays Designed from Ternary Decision Diagrams
• Design of Repairable Cellular Arrays on Multiple-Valued Logic
• On a Class of Multiple-Valued Logic Functions with Truncated Sum, Differential Product and Not Operations
• Design of a Multiple-Valued Cellular Array (Special Issue on Multiple-Valued Integrated Circuits)
• Cortical Dysplasia Detection Method with Support Vector Machine in Pediatric Brain MR Images(International Forum on Medical Imaging in Asia 2009 (IFMIA 2009))
• A New Sulcus Extraction Algorithm Using MAGNET Principle
• Classification for data of hematopoietic tumor patients with fast block-matching-based self-organizing map learning in dynamic environments (特集 医用システム)
• Self-Organizing Map Based Data Detection of Hematopoietic Tumors(Nonlinear Problems)
• Self-Organizing Map Based on Block Learning(Nonlinear Problems)
• A Self-Organizing Map Approach for Detecting Confusion between Blood Samples
• On Fault-Tolerant Fuzzy Controllers Based on Shifting Fuzzy Variables

もっと見る 閉じる

スポンサーリンク