Pattern Synthesis for Nonparametric Pattern Recognition

Parametric methods first choose the form of the model or hypotheses and estimates the necessary parameters from the given dataset. The form, which is chosen, based on experience or domain knowledge, often, need not be the same thing as that which actually exists (Duda, Hart & Stork, 2000). Further, apart from being highly error-prone, this type of methods shows very poor adaptability for dynamically changing datasets. On the other hand, non-parametric pattern recognition methods are attractive because they do not derive any model, but works with the given dataset directly. These methods are highly adaptive for dynamically changing datasets. Two widely used non-parametric pattern recognition methods are (a) the nearest neighbor based classification and (b) the Parzen-Window based density estimation (Duda, Hart & Stork, 2000). Two major problems in applying the non-parametric methods, especially, with large and high dimensional datasets are (a) the high computational requirements and (b) the curse of dimensionality (Duda, Hart & Stork, 2000). Algorithmic improvements, approximate methods can solve the first problem whereas feature selection (Isabelle Guyon & André Elisseeff, 2003), feature extraction (Terabe, Washio, Motoda, Katai & Sawaragi, 2002) and bootstrapping techniques (Efron, 1979; Hamamoto, Uchimura & Tomita, 1997) can tackle the second problem. We propose a novel and unified solution for these problems by deriving a compact and generalized abstraction of the data. By this term, we mean a compact representation of the given patterns from which one can retrieve not only the original patterns but also some artificial patterns. The compactness of the abstraction reduces the computational requirements, and its generalization reduces the curse of dimensionality effect. Pattern synthesis techniques accompanied with compact representations attempt to derive compact and generalized abstractions of the data. These techniques are applied with (a) the nearest neighbor classifier (NNC) which is a popular non-parametric classifier used in many fields including data mining since its conception in the early fifties (Dasarathy, 2002) and (b) the Parzen-Window based density estimation which is a well known non-parametric density estimation method (Duda, Hart & Stork, 2000).