NEWS.md
poisson
in addition to gaussian
and binomial
.s()
.uncertainty_method
argument allows estimation of epistemic uncertainty.predict()
and autoplot()
.validation_split
parameter for monitoring validation losses during training.plot_history()
function for visualizing training/validation loss curves per term and per backfitting iteration.diagnose()
function which provides a 2×2 diagnostic panel.verbose
parameter is now used along all the required functions.
Tensorflow and Keras are now loaded when library(neuralGAM)
is invoked for the first time, and therefore the first run of the neuralGAM()
function has all the required packages ready.