autocompare workflow

Here’s the steps in the workflow for comparing different machine learning models.

Bold terms are defined on the page describing this workflow in detail, linked below this list.

0. Label a small set of songs to provide training data for the models, typically ~20 songs.

1. Pick the models you will compare.

2. Extract features for those models from song files that will be used to train the model.

3. Pick the hyperparameters for the different models.

4. Fit the models to the data

5. Select the best model