partial_fit method?

[graus]

I consider this a point or matter of discussion, debate, or dispute: the implementation of a partial_fit method for incremental learning.

[larsmans]

Ack. Patches are appreciated.

[kmike]

FYI: https://github.com/larsmans/seqlearn/pull/10 aims to implement something similar to partial_fit, but with a very different API. What should partial_fit method do exactly? Should it learn on partial data or should it do a "warm start", or both?

[larsmans]

partial_fit is a method that some scikit-learn estimators have (traditionally SGDClassifier and MiniBatchKMeans, since 0.14 also Naive Bayes). They allow incremental learning: after each partial_fit, you have a functioning estimator/model, which the next partial fit will adjust. (Similar to warm starts.)

[bboalimoe]

hello，can i ask u one question that how to speed up the learning and predicting precess... or can you recommend a production version of supervised hmm sequence learning model~ like mahout's hmm~~~~ thanks a lot!!!