Tribuo

Tribuo is an open source machine learning library written in Java and maintained by Oracle. It gives Java developers a single, strongly typed interface for classification, regression, clustering, anomaly detection and multi label classification, along with tools for model development and evaluation. The library wraps popular third party packages, including XGBoost, LIBLINEAR, LIBSVM and TensorFlow, behind one common API, so new prediction techniques can be swapped in without rewriting application code. It supports the ONNX model exchange format, which lets models trained in Python tools such as scikit learn or PyTorch be deployed inside a Java program, and lets many Tribuo models be exported for use elsewhere. Every model, dataset and evaluation carries provenance recording the parameters, transformations and files used to create it, so a model can be rebuilt exactly and experiments can be traced. Because the API is typed, a loaded model reports what inputs it expects and what labels it can predict, and mistakes are caught at compile time. It also includes built in support for natural language processing tasks such as document classification. Tribuo is aimed at Java and JVM developers who want to build and serve machine learning models without leaving their existing stack. It is released under the Apache 2.0 license and is free to use, distributed through Maven and Gradle and available on GitHub.