LAWS.bend source

LAWS.bend on the hub · documented module

import Baseimport ./src/batch.bend as Batchimport ./src/linear_regression.bend as Linear# Structural guarantees only: these make no claims about F32 accuracy.law prediction_count:  for +model: Linear.Model  for +rows: Batch.Batch<F32>  {Batch.count(F32, Linear.predict_batch(model, rows)) == Batch.count(F32, rows) : Nat}law rejects_empty_training:  {Linear.fit(Batch.Empty{}) == Fail{Linear.NotEnoughSamples{}} : Result<&2, &2, Linear.FitError, Linear.Model>}