Github user myui commented on the issue:
https://github.com/apache/incubator-hivemall/pull/123
Enabling [instance-wise L2 normalization
](https://github.com/apache/incubator-hivemall/pull/123/files#diff-07f782d5891a557e0af0b31638b5680fR179)
resulted in worse performance than `-no-norm`.
Also, training of FFM will become significantly slow when enabling
instance-wise L2 normalization as training does not converge well. More
investigation is required.---
