i want to use fm/ffm in spark pipeline and servering by mleap, anyone has a good solution?
貢獻者指南
技術棧
scala
領域
machine learningdata
議題類型
research
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
5
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
over 1 week
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
stale
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
needs investigation
前置要求
Familiarity with MLeapKnowledge of FM/FFM algorithmsScala experience
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
5
研究方向
This issue asks for a way to use FM (Factorization Machines) or FFM (Field aware Factorization Machines) within the MLeap pipeline and serve via MLeap. Since no solution is provided, the first step is to investigate MLeap's current support for custom transformers and serialization formats. Look at the MLeap source code (e.g., `mleap core/` and `mleap runtime/`) to understand how existing algorithms like linear regression are implemented. Then research FM/FFM implementations in Scala/Spark (e.g., from Spark MLlib or third party libraries) and evaluate the effort to wrap them as MLeap transformers. The maintainers may need to clarify if this falls under their roadmap.
