Using Fisher Information Criteria for Chemical Sensor Selection via Convex Optimization Methods

Abstract

This report presents methodology for the near optimal selection of chemical sensors in a chemical sensing array. While the sensing criteria are task specific, generally one may consider a criterion which maximizes the signal strength or conversely minimizes global error to be best. The quantification of this criteria proceeds from the determinant of the inverse Fisher information matrix which is proportional to the global error volume. If a practitioner has a suitable probabilistic noise model for his or her chemical sensing array and pool of available sensors, the Fisher information matrix may be parametrized to select the best sensors after an optimization procedure. Due to the positive definite nature of the Fisher information matrix, convex optimization may be used to accomplish this task. This report presents the derivation of the supporting set-up, expressions, and constraints for this procedure.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Nov 16, 2016
Accession Number
ADA640843

Entities

People

  • Kevin Johnson
  • Ádám Knapp

Organizations

  • United States Naval Research Laboratory

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Chemical Detection
  • Chemical Detectors
  • Covariance
  • Detection
  • Detectors
  • Estimators
  • Gaussian Noise
  • Inequalities
  • Information Theory
  • Machine Learning
  • Military Research
  • Numbers
  • Probability
  • Probability Density Functions
  • Probability Distributions
  • Random Variables
  • Standards

Readers

  • Adaptive Control and Estimation with Uncertainty in Dynamic Systems.
  • Regression Analysis.
  • Sensor Fusion and Tracking Systems.