nueramic_mathml.ml.metrics.best_threshold#

best_threshold(x, y_true, model, metric='f1', step_size=0.01)[source]#

Returns best threshold by metric by linear search

Parameters:

  • x (Tensor) – training tensor

  • y_true (Tensor) – target tensor. array with true values of binary classification

  • model (Module) – some model that returns a torch tensor with class 1 probabilities using the call: model(x)

  • metric (Literal['f1', 'by_roc']) – name of the target metric that we need to maximize. by_roc - difference between TPR and FPR

  • step_size (float) – step size of linear search

Returns: