Weights for the ensemble model are currently computed via the mean-squared error between the observed and fitted values. However, this method will tend to favor models that overfit the data. Instead, we could implement a weighting scheme based on leave-one-out cross-validation. See the "testing" branch opened by rockclimber112358 for a potential implementation.
Weights for the ensemble model are currently computed via the mean-squared error between the observed and fitted values. However, this method will tend to favor models that overfit the data. Instead, we could implement a weighting scheme based on leave-one-out cross-validation. See the "testing" branch opened by rockclimber112358 for a potential implementation.