This pull request adds --report and --bins command-line arguments in the train-cloud-microphysics app:
--report sets the reporting interval, i.e., the number of epochs between printing the cost function to the screen and writing it to the cost.plt file.
--bins sets the number of phase-space hypercubes across the range of each of the 5 output variables. Only one data point per bin retained for using in training a neural network. This choice maximizes the information entropy for a given retained training data set size.
This pull request adds
--report
and--bins
command-line arguments in the train-cloud-microphysics app:--report
sets the reporting interval, i.e., the number of epochs between printing the cost function to the screen and writing it to thecost.plt
file.--bins
sets the number of phase-space hypercubes across the range of each of the 5 output variables. Only one data point per bin retained for using in training a neural network. This choice maximizes the information entropy for a given retained training data set size.