Increasing the Robustness of Retrieval by Noisy Learning

Abstract

We consider the retrieval behavior of neural network model trained by shaping the energy landscape using noisy version of the patterns. By this process, the spurious states arising during learning is systematically eliminated and the robustness of retrieval is enhanced by noisy learning. Computer simulation for perceivable correlated patterns are shown.

Indexing terms:

• Dilution
• Hopfield model
• Spin glass model

Additional information

Notes on contributors

Vijendran V G

Vijendran Gopalan completed his masters degree in Physics from Madurai Kamaraj University in 1989 and PhD from Center for Spatial Information Processing, Jawaharlal Nehru Technological University, Hyderabad in 1999.

His core skills are image processing, pattern recognition, neural network and OCR. He has developed packages like Khoros Image Processing environment, WIT, Metlab.

He has worked in ISRO, British Telecommunications Plc, London. Presently working as Algorithms Engineer with M/s KLA-Tencor Corporation, California.

He has wide range of publications to his credit. He is a member of IEEE, Man, Cybernetics and Indian User Group.

R Krishnan

R Krishnan was born on December 3, 1949 at Palghat, Kerala. He obtained his BE and MS degrees in Electrical Engineering from Madras University. He obtained PhD from the Indian Institute of Science, Bangalore.

From 1977 till today, he is in the Department of Space at Hyderabad. First he was with the National Remote Sensing Agency and from 1988 he is with the Advanced Data Processing Research Institute (ADRIN), Hyderabad. He was elected as a Fellow of the National Academy of Sciences in 1993. His major contributions are Developing Remote Sensing Data Preprocessing, software for the Indian LANDSAT Earth Station program and Real Time processing software for the Indian' Remote Sensing (IRS) program.

He initiated the Artificial Intelligence Activity at NRSA and developed an Expert System for Geomorphology. He headed a team which developed automatic image understanding methods at NRSA. He played a major role in conceptualizing and in the setting up of the Advanced Data Processing Research Institute (ADRIN). Under his guidance, several software packages in the areas of Digital Mapping, Image Processing and Decision Support Systems have been developed. He has guided 3 PhD students and several MTech students. He has more than 30 research publications to his credit. He was invited by the International Atomic Energy Agency, Vienna as an Expert to advise them on the use of Remote Sensing Techniques.