gpytorch.metrics

See the metrics tutorial for usage instructions.

metrics.mean_absolute_error(test_y)

Mean absolute error.

Parameters:
metrics.mean_squared_error(test_y, squared=True)

Mean squared error.

Parameters:
metrics.mean_standardized_log_loss(test_y, train_y=None)

Mean standardized log loss.

Computes the average standardized log loss, which subtracts the loss obtained under the trivial model which predicts with the mean and variance of the training data from the mean log loss. See p.23 of Rasmussen and Williams (2006).

If no training data is supplied, the mean log loss is computed.

Parameters:
metrics.negative_log_predictive_density(test_y)

Negative log predictive density.

Computes the negative predictive log density normalized by the size of the test data.

Parameters:
metrics.quantile_coverage_error(test_y, quantile=95.0)

Quantile coverage error.

Parameters: