gilbcharlene edited this page Jun 5, 2012 · 5 revisions


ECL-ML has a document that serves as a walk-thru and reference manual. This Wiki is a supplement to that and serves to give more immediate and supplemental information


ECL-ML is still under early development; the first alpha release is available in branch 1.0.0rc. All of the routines have undergone testing and any known issues are noted here. A document is available describing this library; available from the HPCCSYSTEMS.COM website. Please note - this is ALPHA - NONE of the interfaces are considered sacrosanct. Future releases MAY NOT be backwards compatible with the current code.


ML is the main directory for the machine learning libraries; the routines in there should be considered as the interface to ML. It currently also contains two subservient libraries: Mat (Matrices) and Docs (Documents). It is possible that one or both of those will eventually be pulled out; so using IMPORT ML.Mat as Mat; is recommended.


Under the Tests library, you will find several sub tests folders:

Explanatory: mirrors the documentation (mainly small tests)

Benchmarks: shows performance and act as system validation (larger tests)

Validation: places where our code is measured against other systems (Mahout &c)

Deprecated: old tests, to be removed at a later stage


This wiki contains a Function Reference page that includes detailed descriptions for most ECL-ML functions.