Version v.0.7.1
This repository contains four libraries for working with PFA (see http://scoringengine.org for details).
- Hadrian: complete implementation of PFA for the JVM (Java, Scala, ...)
- loads, validates, byte-compiles, and executes PFA scoring engines within a JVM
- tools for analyzing PFA statically and at runtime
- Titus: complete implementation of PFA for Python
- loads, validates, interprets, and executes PFA scoring engines in pure Python
- translates PrettyPFA, Python code, and some PMML to PFA
- produces models using k-means and CART
- interactive analysis of PFA documents with the PFA Inspector script
- PFA development tools, such as pfachain and pfaexternalize
- Aurelius: tools for building PFA from R
- converts a subset of the R language to PFA
- converts glm, glmnet, randomForest, and gbm models to PFA
- integrated with Titus for PFA validation and execution
- Antinous: model producer in Jython
- encapsulates arbitrary Jython code in a PFA container
- produces models using k-means
It also contains four envelopes for embedding PFA.
- hadrian-standalone: runs multithreaded PFA on a command line
- quick, simple testing in full Hadrian (as opposed to quicker testing in Titus)
- may be used as a component of a shell-based workflow
- hadrian-mr: runs PFA in Hadoop
- schema-matching and sanity checks before job execution
- built-in secondary sort semantics
- batch scoring with or without snapshot-as-output
- may be used with Antinous to produce models in Hadoop
- hadrian-gae: runs Hadrian on Google App Engine
- PFA-as-a-service
- backend for tutorial examples on http://scoringengine.org
- hadrian-actors: complex directed acyclic workflows for PFA
- actor-based framework capable of complex, multi-engine workflow topologies
- extensible configuration language
They are all licensed under the Hadrian Personal Use and Evaluation License (PUEL).