[HIVEMALL-24] Fix the prediction logic of Field-aware Factorization Machines more scalable

• Type: Improvement
• Status: Closed
• Priority: Critical
• Resolution: Fixed
• Fix Version/s: 0.5.2
• Assignee: Makoto Yui
• Reporter: Makoto Yui
• Created: 16/Jan/17 07:27
• Updated: 04/Sep/18 07:10
• Resolved: 04/Sep/18 07:10

Description

We already introduced a draft implementation of Field-aware Factorization Machines [2] in [1] but it's prediction scheme is not scalable.

Need to fix the prediction logic for production uses.

[1] https://github.com/myui/hivemall/pull/284
[2] http://www.csie.ntu.edu.tw/~cjlin/papers/ffm.pdf

Issue Links

• blocks: HIVEMALL-152 [Umbrella] Release Hivemall v0.5.2
• relates to: HIVEMALL-154 Refactor Field-aware Factorization Machines to support Instance-wise L2 normalization
• links to: GitHub Pull Request #58
• links to: GitHub Pull Request #105