train.in and cross.in

The files train.in and cross.in contain the training and cross-validation data, respectively. The first line contains an integer N with the total number of entries to be read in. There are two different ways of interpreting the data, depending on how flag_ts (the first parameter in data.in) has been set.

Time series prediction: flag_ts=1
A time series of length N is read in.

Interpolation: flag_ts=0
N must be devideable by input_grid_size+1, where input_grid_size is the size of the input grid, defined in topo.in. The first input_grid_size numbers are interpreted as the input elements (presented at the input layer), while the following value is defined as the target variables. The next input_grid_size entries are again interpreted as input elements, followed by another target variable. This scheme is repeated throughout the whole file.

The cross-validation set is only read in if flag_cross=1 (where flag_cross is the second element in data.in).


Last modified: May 2000