The documentation offers several entry points to the AutoML framework:
- User-callable functions Use the interface with no need for detailed information on implementation
- Low-level introduction to procedures undertaken within the framework; sections correspond to the core elements of most machine-learning workflows:
- Data pre-processing (data retrieval, data cleaning, feature extraction)
- Data processing (model selection and optimization)
- Data post-processing (saving reports, models, graphs and metadata)
- Advanced parameter modifications Change underlying functionality using tunable parameters
- FAQ
Version 0.3.0 of the AutoML framework has undergone fundamental changes with respect to the coding structure. In particular, the framework has moved from a small number of closely-dependent functions to a coding pattern which separates the individual pieces of required functionality into distinct sections. This is facilitated by the directed acyclic graph structure.
Understanding the structure provides insights into the functionality and their interdependencies. It also explains the documentation breakdown within the data pre-processing, data processing and data post-processing sections, which reference the applied functions within each of these sections based on their node within the graph i.e.
.automl.trainTestSplit.node.function
.automl.saveGraph.node.function
.automl.featureData.node.functionInterconnections between sections of the framework:
