Featureless pattern recognition in an imaginary Hilbert space

The featureless methodology is applied to the class of pattern recognition problems in which the adopted pairwise similarity measure possesses the most fundamental property of inner product to form a nonnegative definite matrix for any finite assembly of objects. It is proposed to treat the set of all feasible objects of recognition as a subset of isolated points in an imaginary Hilbert space. This idea is applied to the problem of determining the membership of a protein given by its amino acid sequence (primary structure) in one of preset fold classes (spatial structure) on the basis of measuring the likelihood that two proteins have the same evolutionary origin by way of calculating the so-called alignment score between two amino acid sequences, as it is commonly adopted in computational biology.