Compare vikos to vowpal wabbit

[MaxBenChrist]

At the moment, vikos is fitting its models by online algorithms, one event at a time. This is quite similar to vowpal wabbit.

Vowpal Wabbit (VW) project is a fast out-of-core learning system sponsored by Microsoft Research and (previously) Yahoo! Research.
Source

Vowpal wabbit is also fitting Logistic or Linear Regression Models. It would be interesting to compare both projects in terms of speed or design decisions.

[pacman82]

I did not find the time for an in depth comparison yet, however here are my findings from the first quick look:

Commonalities:

• Cost function is changeable
• Model is changeable
• SGD based training algorithm (more sophisticated than current implementations in vikos)

Differences:

• Is simply more mature in terms of features
• provides functionality for regularization
• It's generic capabilities are far inferior to vikos. Since it's written in C++ which forces it to make a lot more decisions at runtime to still be usable without type inference. However, because of this it can also provide a really handy command line interface.
• cost function for binary classifiers rather use -1..1 instead of 0..1 used by Vikos. I can not say if this choice is arbitrary or has a significant advantage. I choose 0..1 for its easy interpretation as a probability.