Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

Fusion rule estimation using vector space methods

Conference ·
OSTI ID:471398
In a system of N sensors, the sensor S{sub j}, j = 1, 2 .... N, outputs Y{sup (j)} {element_of} {Re}, according to an unknown probability distribution P{sub (Y(j) /X)}, corresponding to input X {element_of} [0, 1]. A training n-sample (X{sub 1}, Y{sub 1}), (X{sub 2}, Y{sub 2}), ..., (X{sub n}, Y{sub n}) is given where Y{sub i} = (Y{sub i}{sup (1)}, Y{sub i}{sup (2)}, . . . , Y{sub i}{sup N}) such that Y{sub i}{sup (j)} is the output of S{sub j} in response to input X{sub i}. The problem is to estimate a fusion rule f : {Re}{sup N} {r_arrow} [0, 1], based on the sample, such that the expected square error is minimized over a family of functions Y that constitute a vector space. The function f* that minimizes the expected error cannot be computed since the underlying densities are unknown, and only an approximation f to f* is feasible. We estimate the sample size sufficient to ensure that f provides a close approximation to f* with a high probability. The advantages of vector space methods are two-fold: (a) the sample size estimate is a simple function of the dimensionality of F, and (b) the estimate f can be easily computed by well-known least square methods in polynomial time. The results are applicable to the classical potential function methods and also (to a recently proposed) special class of sigmoidal feedforward neural networks.
Research Organization:
Oak Ridge National Lab., TN (United States)
Sponsoring Organization:
USDOE Office of Energy Research, Washington, DC (United States)
DOE Contract Number:
AC05-96OR22464
OSTI ID:
471398
Report Number(s):
CONF-970465--18; ON: DE97006006
Country of Publication:
United States
Language:
English