ibayer / fastFM

fastFM: A Library for Factorization Machines
http://ibayer.github.io/fastFM
Other
1.07k stars 206 forks source link

New Algorithm of Solving FM #73

Open wschin opened 7 years ago

wschin commented 7 years ago

According to our recent study, Newton method can also be applied to solve FM as well. I am not sure if this result is helpful to fastFM, but if you are interested in how it works. Please see http://www.csie.ntu.edu.tw/∼cjlin/papers/fm.

Wei-Sheng

ibayer commented 7 years ago

Link doesn't work.

wschin commented 7 years ago

It seems that "~" is replaced by another special character after copy-and-paste. Please fix it manually. Thanks.

ibayer commented 7 years ago

I see, here is the proper formatted link. An Efficient Alternating Newton Method for Learning Factorization Machines Thanks for the heads up the paper looks interesting. I wanted to replace the coordinate descent / als solver for classification for a while. It uses the probit link function which is not as stable as logit. However, I'm not sure if I'll get to it any time soon.

mblondel commented 7 years ago

@wschin Your paper is very interesting. We would welcome a contribution in polylearn.

Why did you decide to call the model considered in your paper "factorization machine"? In my ICML paper, I called it polynomial network.

wschin commented 7 years ago

@mblondel: Really appreciate your invitation, but I am blocked by other projects. For the name, we followed Rendle when initiating this work (1 year ago), so... Anyway, I will keep your comment in mind. Thanks a lot.