search

VinceShieh / spark-ffm

FFM (Field-Awared Factorization Machine) on Spark
Apache License 2.0
105 stars 44 forks source link

features to complete #2

Open VinceShieh opened 7 years ago

VinceShieh commented 7 years ago

I summarize the current design and come out with a few things that can be done to make spark-ffm better.

  1. add global bias intercept and one-way interactions, current design only support pairwise interactions
  2. normalization support

It'd be nice if any of you are interested in working on this project, you are always welcome and I will try best to help. :)

lmxiaonuo commented 7 years ago

hello, I have used your ffm to train some model in big data, but the tr_loss turns out to be NaN when the data size is larger than 1000000, please do a favor

Thinkgamer commented 6 years ago

@lmxiaonuo I have also encountered this problem. Have you solved it?

VinceShieh commented 6 years ago

I just made a PR for this issue, you can get it from https://github.com/VinceShieh/spark-ffm/pull/16 and test on your end. And also please open an issue if the problem still exists.

Thinkgamer commented 6 years ago

@VinceShieh I have sloved it! thank you

VinceShieh commented 6 years ago

@Thinkgamer did u resolve the issue with the PR I submitted or some other means you tried? If the PR works, I will merge it. Thanks.

Thinkgamer commented 6 years ago

@VinceShieh I did not try your PR，I did some processing on the features. Next I will test your changes and then send feedback. Thank you.