Hi devs, Yun and I would like to start a discussion for releasing Flink ML <https://github.com/apache/flink-ml> 2.0.0
In the past few months, we have made a substantial amount of improvements in Flink ML. These improvements include FLIP-173 <https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=184615783>, FLIP-174 <https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=181311361>, FLIP-175 <https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=181311363>, FLIP-176 <https://cwiki.apache.org/confluence/pages/viewpage.action?pageId=184615300>, and a few representative algorithms (e.g. KMeans <https://github.com/apache/flink-ml/pull/27>, NaiveBayes <https://github.com/apache/flink-ml/pull/32>, LogisticRegression <https://github.com/apache/flink-ml/pull/28>, KNN <https://github.com/apache/flink-ml/pull/24> and OneHotEncoder <https://github.com/apache/flink-ml/pull/37>). And we plan to introduce these improvements in FFA 2021 <https://flink-forward.org.cn/#agenda>. We are hoping to complete the official release for Flink ML before the FFA videos are released in January. A few months ago we moved the flink-ml module from the core Flink repo to the flink-ml repo. This allows us to decouple the flink-ml release from the core Flink release. Prior to that change, the flink-ml module was released as part of the core Flink repo. Its latest version in maven is 1.8.3 (flink <https://mvnrepository.com/artifact/org.apache.flink/flink-ml>). Given the amount of changes we have made in this release, it seems reasonable to make a major version bump and use 2.0.0 as the next Flink ML version. Note that the upcoming Flink ML release depends on Flink 1.14.x as the minimal Flink version. If there is any concern, please let us know. Cheers, Dong
