The DATASET on which the performance has to be measured.
RETURNS
1.0 - (cs/ts) where cs is the number of correctly classified samples
and ts is the total number of samples, or -1.0 if an error occured.
FUNCTION
This routine first checks whether the network is correct: NetFlag should be
MLNET (see NET-flags); NumOutputs should be at least 2
(i.e., at least 2 classes) and NumInputs should match dset->NumInputs.
The network is evaluated (using eval_ff_net) for every SAMPLE in
dset. The output unit having the maximum output value is considered to be the network's classification
output. If the InOutIndex of that unit matches the Output.Label of the
sample, the sample is judged correctly classified. The return value is calculated
as stated above.
