Alink is a machine learning platform built by Alibaba's computing team. It provides a large library of ready-to-use machine learning algorithms that can run either on a batch of stored data or on a continuous stream of incoming data. The platform is built on top of Flink, a system for processing large amounts of data quickly, so it can handle both one-time analysis jobs and always-on streaming jobs. The library covers a broad range of algorithm types: classification, clustering, data mining, feature engineering, recommendation, and more. The algorithms are packaged as components that you connect together into pipelines, similar to assembling building blocks. You configure each component by setting parameters, then link them in sequence to define the full processing flow. Alink can be used from Java or from Python. The Python interface is called PyAlink and installs via pip. It works with Jupyter Notebook, which is an interactive coding environment that runs in a browser. The Python and Java APIs are designed to mirror each other, so the way you configure and connect algorithm components looks similar in both languages. Installation requires Python 3.6, 3.7, or 3.8 and Java 8. The specific package you install depends on which version of Flink your setup uses, PyAlink supports Flink versions 1.9 through 1.13. For Java projects, you add the Alink library as a dependency using a standard build tool. The project includes tutorials in both Java and Python versions, along with documentation listing all available algorithm components. It was developed and is maintained by Alibaba's PAI (Platform for AI) team.
← alibaba on gitmyhub — every repo by this author, as a profile.
Verify against the repo before relying on details.