!
or #
character. An example must have two lines, the input pattern line
starting with i
or I
, and the output pattern line starting with o
or O
.
When you load a training set file from a model panel, INNE expects a training set
compatible with the connected net; this means that each input pattern must consist of a
number of value at least equal to the number of input unit, and so for each output pattern
too; every value in advance will be ignored. A training set that doesn't respect the
previous requisites will not be loaded.
As an example, we give the training set file for the XOR boolean function. The function
has to give 0 in output whenever the input is 0 0 or 1 1; while it has to output 1 when
the input is either 0 1 or 1 0. With two input nodes there are no more examples to
submit to a network computing the XOR function. The corresponding training set file
could be the following. Commments are introduced by the #
character.
#Training Set for the XOR function: # Example 1, input I 0 0 # output O 0 # Example 2, input I 0 1 # output O 1 # Example 3, input I 1 0 # output O 1 # Example 4, input I 1 1 # output O 0