Bugs in org.encog.neural.pattern
Most comments preceding generate() method in patterns are plain wrong. For example, FeedFowardPattern's generate() method has comment that says return value is the Elman neural network. Beside comments, FeedForwardPattern's genrate() method will return a network with SimpleRecurrentLogic(). Shouldn't it return FeedForwardLogic?




The comments may suffer from a few copy/paste errors. Thanks for the note. I will take a look.
As to the feed forward logic, the srn class is the most flexable as it can handle feedforward, as well as srn. That is why it iss the default. If you want to force feedford logic only, you can use the feedforward logic only class, it is slightly faster. But it is not really in correct to use srn. The code just uses the original default class.
It seems that feedforwardlogic is the only logic that is failing with GA. It complains about about unknown source. I tried with different logic and everything works fine.
I think if you add a call to finalizeStructure after you change the logic, things will work better.