EXTENDED DEPENDENCY ANALYSIS OF LARGE SYSTEMS∗

Publisher Website
Google Scholar
Abstract
Probabilistic Reconstructability Analysis and Dependency Analysis are useful tools for the detection and analysis of structure in multivariate systems, but heretofore they have been limited to systems with few variables, for reasons discussed in the paper. The Extended Dependency Analysis described here uses heuristic methods to overcome the combinatorial constraints which limit the earlier techniques. The method is illustrated and the results of substantial testing are given. Each test involved a randomly-generated network of up to 100 nearly-deterministic automata, each one connected to up to five others in the network; from data generated by the network, the Extended Dependency Analysis technique deduced the structure of the network as well as the automaton mappings. The test results indicate that the method is nearly error-free

This publication has 10 references indexed in Scilit: