Optimal informational sorting: The ACS-ULA approach

Abstract

This paper introduces a new methodology for optimal informational sorting that is based on the combination of two neural network algorithms: Activation and Competition System (ACS) and the Universe Lines Algorithm (ULA). We present the basic motivation and the technical details of the methodology and carry out a benchmark based upon the classical West Side Story database originally introduced by McClelland. In the benchmark, two simple alternative methodologies are tested: Linear Correlation (LC) and Prior Probability (PP) and it is found that both provide a commonly biased sorting of the database. If we provide the ACS-ULA methodology the output of LC and PP as the only input, it turns out that ACS-ULA carries out the sorting flawlessly and moreover constructs a structural characterization of the database that is entirely different than the one provided by the two input methodologies. This suggests that the ACS-ULA methodology might establish a new benchmark for optimal information sorting and structural analysis of databases.