If the user requires that the machine learning model must be compliant with certain specifications, Silas provides a way to build optimised models that are guaranteed to meet the specifications.
For example, to build a model using settings1.json and enforce specs in specs1.json and output the model in model/this-model/, issue the following command:
silas learn -e specs1.json -o model/this-model/ settings1.json
The user can verify the correctness of the generated models using the commands in Verifying The Model Against User Specifications. However, this is not necessary as the generated models should be correct by construction.