Simple Bayesian Classifiers Do Not Assume Independence

Here we show that the SBC is in fact optimal even when the independence assumption is grossly violated, and thus applicable to a far broader range of domains than previously thought.