Extraction of physical laws from joint experimental data
Igor Grabec update to 2007-10-10
The extraction of a physical law y=yo(x) from joint experimental data about xand y is treated. The joint, the marginal and the conditional probabilitydensity functions (PDF) are expressed by given data over an estimator whosekernel is the instrument scattering function. As an optimal estimator of yo(x)the conditional average is proposed. The analysis of its properties is basedupon a new definition of prediction quality. The joint experimental informationand the redundancy of joint measurements are expressed by the relative entropy.With the number of experiments the redundancy on average increases, while theexperimental information converges to a certain limit value. The differencebetween this limit value and the experimental information at a finite number ofdata represents the discrepancy between the experimentally determined and thetrue properties of the phenomenon. The sum of the discrepancy measure and theredundancy is utilized as a cost function. By its minimum a reasonable numberof data for the extraction of the law yo(x) is specified. The mutualinformation is defined by the marginal and the conditional PDFs of thevariables. The ratio between mutual information and marginal information isused to indicate which variable is the independent one. The properties of theintroduced statistics are demonstrated on deterministically and randomlyrelated variables.
journal: Eur. Phys. J. - B, 48, 279-289 (2005)