Adaptive Stochastic Resonance and Noise Processing

University Of Southern California, Los Angeles CA

Investigators

Abstract

0070284 Kosko This project will explore how noise can improve signal processing and computation both in theory and in select information and physical systems. The launching point for this project is the principal investigator's recent publication that shows that many nonlinear dynamical systems can adoptively achieve a noise-based "stochastic resonance" where external noise improves the system's signal-to-noise ratio or other performance measure (such as measures of cross-correlation, mutual information, or probability of detection). The project will more generally explore "noise processing" and how it affects many feedback and some feedforward systems. The central focus will be the study of adaptive stochastic resonance (ASR). ASR uses sample data and as few statistical assumptions as possible to find the optimal type or amount of noise to add to a nonlinear system to improve some system performance measure such as the system's spectral signal-to-noise ratio. One may view noise simply as an unwanted signal. But noise itself is a free and local source of energy. So noise can sometimes help computation as well as interfere with it. A key part of the initial effort will focus on how to achieve ASR in feedback systems that try to detect broadband forcing signals. This problem has so far defied solution. The Project Description lists five core problems with subproblems that this research effort will try to solve. The work plan presents the PI's current best estimates of how to tackle these problems. The effort will involve a constant search for real-world applications of this emerging field of adaptive noise processing.