Resampling data on a time axis in tasks of revealing regularities
A. Vinogradov · Pattern Recognition and Image Analysis · 2008
A framework is proposed for revealing and sharpening discriminative regularities in data. Noninvariance of the data sample with respect to transformations of this or that type is used as indicator of the presence of some useful regularity or dependency. Weak regularities are emphasized by deletion from the time axis of all results of consecutive trials which form maximal invariant subset.