An Empirical Comparison of ICA Over Finite Fields Algorithms

A series of algorithms to perform ICA over Finite Fields have been proposed in recent years. They differ from each other in terms of search strategy and independence criterion, which may yield different performances. Hence, this work intends to compare the two most relevant heuristic-based approaches, namely AMERICA and MEXICO, with the recently developed cobICA, which is based on a state-of-the-art immune inspired optimization technique. Experimental results indicate AMERICA with the best separation quality, while cobICA presents intermediate results but with a tendency of smaller computational cost in high dimension scenarios.