MLJBase
Repository for developers that provides core functionality for the MLJ machine learning framework.
| Branch | Julia | Build | Coverage |
|---|---|---|---|
master |
v1 |  |
 |
dev |
v1 |  |
 |
MLJ is a Julia framework for combining and tuning machine learning models. This repository provides core functionality for MLJ, including:
-
completing the functionality for methods defined "minimally" in MLJ's light-weight model interface MLJModelInterface (/src/interface)
-
definition of machines and their associated methods, such as
fit!andpredict/transform(src/machines). Serialization of machines, however, now lives in MLJSerialization. -
MLJ's model composition interface, including learning networks, pipelines, stacks, target transforms (/src/composition)
-
basic utilities for manipulating datasets and for synthesizing datasets (src/data)
-
a small interface for resampling strategies and implementations, including
CV(),StratifiedCVandHoldout(src/resampling.jl) -
methods for performance evaluation, based on those resampling strategies (src/resampling.jl)
-
one-dimensional hyperparameter range types, constructors and associated methods, for use with MLJTuning (src/hyperparam)
-
a small interface for performance measures (losses and scores), implementation of about 60 such measures, including integration of the LossFunctions.jl library (src/measures). To be migrated into separate package in the near future.