*********** Sums.nns is a NeuNet Pro Sample File ***********
This file contains special properties that allow NeuNet Pro to recognize it
as authorized NeuNet sample data. Anyone using the unlicensed version
of NeuNet Pro is welcome to experiment with this sample data.
Please do not modify this file, or it will lose its status as authorized sample data.
For further information about Neunet Pro and additional sample data,
please visit the NeuNet Pro website at http://www.cormactech.com/neunet
All of this data has been generated by CorMac Technologies Inc.
The provided NeuNet project attempts to train a neural network
to learn to add three integers.
*************** More about the Sums Data *******************
S U M S D A T A
(Train a neural net to add three integers)
This project contains four columns and 10,000 rows.
[Index], [X], [Y], [Z], [SumXYZ]
X,Y,Z are integers randomized between 1 and 1000.
NeuNet Pro was trained on the first 500 rows then tested on
the remaining 9,500 rows. Accuracy was 99.75% on the training
set and 99.74% on the test set.
This data is an excellent way to learn about Neunet Pro.
Remember not to include the row index as one of the inputs.
Notice how the predictions toward the edges of the data space
are not as accurate as those near the center.
Try experimenting with the user MIN/MAX setttings in the
advanced configuration.
Try reducing the size of the training set below 100 rows
and observe how the accuracy on the test set begins to
drop off.
Try experimenting with the number of hidden nodes in the
advanced configuration.