Small Learn Models

A small Learn model exhibits only a simple exponential growth as the number of features increases. The size of the model grows slower with each added model feature. This results in much smaller model files for the same number of features, and thus allows more features to be supported, given the same limitations on the available memory. The accuracy of a trained small Learn model is comparable with the accuracy of the large model type for most datasets, although in rare cases a large model might classify some situations better.

To reduce model training time, dynamic subset training is automatically switched off starting at 17 features. Therefore, a 17-feature model is expected to train much faster than a 16-feature model.