capturing the training set in a closure to avoid retraining for each …

[alexzubiaga]

I know that it's been a while since you updated this repo, but It's helped me so much to grasp ML concepts that I had to contribute back. In my experiments with the Bayes clasificator I noticed very fast training times, but extremely slow clasification times. As it turned out due to the lazy Seq evaluation I was retraining the whole model for each entry from my test set. This can be avoided by capturing the model in a closure.

Many Thanks for this great code!