This keyword indicate number of input (independent) variables.
The output (dependent) variables are defined by keyword OUTPUTS.
Each variable is represented as a column of data table. The input variables
can be before or after output (target) variables depending on REVERSED
keyword. The keyword NAMES indicates that each
data entry contains name in the first column.
The input variables are pre-processed before neural network calculations
in several steps:
- variables indicated in AVOID or not indicated
in INCLUDE are eliminated;
- highly correlated variables, with level of correlation defined by CORRELATION
keyword, are eliminated;
- each input variable is normalized according to range of activation function
of the first layer neurons. The default range is [0.1,0.9] for logistic
function, y=1./(1.+exp(-x)), that is used for all layers.