home *** CD-ROM | disk | FTP | other *** search
- //
- // file: demo2trn.rsp
- // desc: demo2 training response file
- // by: patrick ko shu pui
- // date: 27 apr 1992
- //
- // comments:
- //
- // This resp file is used to train a 2 hidden layer neural net to
- // learn an shift function. The NN has a 4-unit input, 4-unit
- // output, a 4-unit 1st hidden layer, and a 3-unit 2nd hidden layer.
- // The result file is contained in demo2.out where you may view
- // the output. If you output is close to 1 you may interpret it as
- // a 1, and 0 if it is close to 0.
- //
-
- // 4 input units
- -i=4
-
- // 4 output units
- -o=4
-
- // 2 hidden layers
- -hh=2
-
- // 1st hidden layer = 4 units
- -h=4
-
- // 2nd hidden layer = 3 units
- -h=3
-
- // use demo2.trn as training file
- -ftrain=demo2.trn
-
- // and there are 4 training patterns
- -samp=4
-
- // weight dump file is demo2.dmp
- -fdump=demo2.dmp
-
- // report training status every 10 cycles
- -r=10
-
- //
- // the above option in this response file is equivalent to the following command line
- //
- // bptrain -i=4 -o=4 -hh=2 -h=4 -h=3 -ftrain=demo2.trn -samp=4 -fdump=demo2.dmp -r=10
- //
- // and you should use the following command line for recognition
- //
- // bprecog -i=4 -o=4 -hh=2 -h=4 -h=3 -frecog=demo2.rgn -samp=2 -fdump=demo2.dmp -fout=demo2.out
- //
-