Hi devs, Zhipeng and I would like to start a discussion regarding the release of Flink ML 2.2.0. And I would like to volunteer as the release manager.
Over the past few months, we've been focused on enhancing Flink ML's feature engineering capabilities. We're happy to report that the library now includes 33 feature engineering algorithms, covering 28 out of the 33 feature engineering algorithms provided in Spark ML. Here are some highlighted improvements we've made since the last release: - Added 27 new feature engineering algorithms. - Introduced APIs and infrastructure for online serving via FLIP-289, allowing you to serve models online. To start, we've provided the LogisticRegressionModelServable to serve the logistic regression model online, and we'll continue to add more servables in the future. - Added Python support for every algorithm in Flink ML, which means you can run every algorithm using Python. - Added two online algorithms (AgglomerativeClustering and OnlineStandardScaler) which have been deployed in production to classify error logs in real-time. These algorithms will soon be integrated into the open-source project alibaba/SREWorks. With the above improvements, we believe it is time to release Flink ML 2.2.0. We hope that these new features and improvements will assist Flink users in their machine learning tasks. Please feel free to provide your feedback. Best Regards, Dong
