Making LSI evaluatable

[ibnesayeed]

As per the #142 implementation, in order to evaluate a classifier, the classifier instance needs to support a uniform interface of train, classify, categories and reset methods. I have recently added reset functionality to all classifiers #143 as it could be useful even when not evaluated.

Is it OK if I add train, classify, categories, and perhaps untrain method as well in the LSI while internally utilizing existing methods and not breaking any backwards compatibility?

As far as I understood LSI, in order to get the list of all known categories, I will have to scan all the indexed items. Is there any other faster way to get that information?

[Ch4s3]

Yeah that should be fine

[ibnesayeed]

@Ch4s3: Yeah that should be fine

We will take care of it in a separate PR once #142 and #143 are merged.

[Ch4s3]

You should be good to go now. Sorry to keep you waiting.