Revised Draft ML API.

[PondiB]

• Shifted from the STAC ML extension to MLM STAC extension
• Change from "predictor" and "target" to training_set and the "target" parameter can be a column name of the dependent variable

[m-mohr]

@PondiB The subtype doesn't need to follow the STAC naming, so it can still be ml-model instead of mlm-model.

[jonas-hurst]

@PondiB The subtype doesn't need to follow the STAC naming, so it can still be ml-model instead of mlm-model.

Agreed! mlm is short for machine learning model, so saying mlm-model is redundant

[jonas-hurst]

From this description, it is unclear to me how training data (predictors+labels) are passed into the process. The description on the two parameters needs to be clearer.

An alternative approach could be to include the training labels in the raster data cube. This would require a vector-to-raster conversion if training data exists as vector data.

[PondiB]

@PondiB The subtype doesn't need to follow the STAC naming, so it can still be ml-model instead of mlm-model.

Agreed! mlm is short for machine learning model, so saying mlm-model is redundant

Sure