gaussianLoss: Custom loss function for Gaussian distributions.

In SantanderMetGroup/downscaleR.keras: A climate4R package for statistical downscaling <http://meteo.unican.es/climate4r>

Custom loss function for Gaussian distributions.

Description

This loss function optimizes the negative log-likelihood of a Gaussian distribution. It is a custom loss function defined according to keras functions.

Usage

gaussianLoss(last.connection = NULL)

Arguments

last.connection

A string with values "conv" o "dense" depending on the type of the net's last connection. DEFAULT is NULL.

Details

Note that infering a conditional gaussian distribution means to estimate their associated parameters: mean, and variance. To avoid computational instabilities we actually estimate log var. Therefore, make sure that your output layers are designed according to this property of the distribution.

Author(s)

J. Bano-Medina

See Also

bernouilliGammaLoss for computing the negative log-likelihood of a Bernouilli-Gamma distribution downscaleTrain.keras for training a downscaling deep model with keras downscalePredict.keras for predicting with a keras model prepareNewData.keras for predictor preparation with new (test) data downscaleR.keras Wiki downscaleR Wiki for downscaling seasonal forecasting and climate projections.

SantanderMetGroup/downscaleR.keras documentation built on July 7, 2023, 1:22 p.m.