Statistics derived from trained models
type t = {n_samples : int;Number of samples used for training
*)target_variance : float;Variance of targets
*)sse : float;Sum of squared errors
*)mse : float;Mean sum of squared errors
*)rmse : float;Root mean sum of squared errors
*)smse : float;Standardized mean squared error
*)msll : float;Mean standardized log loss
*)mad : float;Mean absolute deviation
*)maxad : float;Maximum absolute deviation
*)}Type of full statistics
val calc_n_samples : Trained.t -> intcalc_n_samples trained
number of samples used for training trained.
val calc_target_variance : Trained.t -> floatcalc_target_variance trained
variance of targets used for training trained.
val calc_sse : Trained.t -> floatcalc_sse trained
the sum of squared errors of the trained model.
val calc_mse : Trained.t -> floatcalc_mse trained
the mean sum of squared errors of the trained model.
val calc_rmse : Trained.t -> floatcalc_sse trained
the root of the mean sum of squared errors of the trained model.
val calc_smse : Trained.t -> floatcalc_smse trained
the standardized mean squared error of the trained model. This is equivalent to the mean squared error divided by the target variance.
val calc_msll : Trained.t -> floatcalc_msll trained
the mean standardized log loss. This is equivalent to subtracting the log evidence of the trained model from the log evidence of a normal distribution fit to the targets, and dividing the result by the number of samples.
val calc_mad : Trained.t -> floatcalc_mad trained
the mean absolute deviation of the trained model.
val calc_maxad : Trained.t -> floatcalc_mad trained
the maximum absolute deviation of the trained model.