An implementation of relational table classification and compression
This work presents some methods for relational table classification and compression from the view points of information theory and approximation theory. Some useful and important definitions and formulas for implementing these methods are defined. An algorithm to partition a relational table into a collection of equivalence classes and then reduce it to a compressed table by using the method of least squares is presented. The measurable levels of the distortion of the final compressed table are analyzed. The choice of the class limiting fraction which influences the construction of the equivalence classes and the choice of the attributes for computing the similarity between two tuples are discussed. An abstract example and examples in scene content analysis are presented to demonstrate the algorithm. A computer implementation of the classification and compression algorithm is also presented.
Thesis78.L555.pdf
1.91 MB
Adobe PDF
e7de7fafe9d471a07e455a4b67984627