model_type parameter specifies the type of model to use.
The following model types are supported:
ecd(default): Encoder-Combiner-Decoder neural network model.
gbm: Gradient Boosting Machine tree-based model.
Every model type has trainers associated with it. See the Trainer section for details about the supported training algorithms per model type.
Model Type: ECD¶
See the ECD documentation for details about the Encoder-Combiner-Decoder deep learning architecture.
The full breadth of Ludwig functionality is available for the
ecd model type.
Model Type: GBM¶
The GBM model type is a gradient boosting machine (GBM) tree model. It is a tree model that is trained using a supported tree learner. Currently, the only supported tree learner is LightGBM.
gbm model type introduces the following limitations:
- only binary, category, and number features are supported
- only a single output feature is supported
combinersection is ignored