home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
HPAVC
/
HPAVC CD-ROM.iso
/
NEUCLS3.ZIP
/
NURN.ZP
/
INT.HLP
< prev
next >
Wrap
Text File
|
1994-09-03
|
2KB
|
50 lines
Fast Training Program;
1. Purpose;
a. Initialize a MLP using random initial weights
b. Train a MLP network using a method much faster than backpropagation
2. Features;
a. Trains network much faster than BP
b. Uses a batching approach, so the order of training
patterns is unimportant
c. Has adaptation of learning factor
d. Shows the training MSE and error percentages
e. Does not save weights to disk, in the demo version
3. Example Run of Fast Training Program
a. Go to the "Batch Processing" option and press <ret>
b. Observe the parameter file with commented keyboard responses;
15, 4. ! number of training iterations, error % threshold
2 ! Enter 1 for existing weights, 2 for new random weights
grng.top ! file storing network structure
1 ! 1 if data file has desired outputs, 0 else
2 ! 1 to choose coded outputs, 2 for uncoded outputs
GRNG ! filename for training data
0 ! # of training patterns (0 for all)
1, 3 ! Enter numbers of 1st and last patterns to examine (0 0 for none)
grng1.wts ! filename for saving the weights
.01 ! learning factor
4 ! 1 to continue, 2 for new network, 3 for new data, 4 to stop
The program will read all patterns from the file grng, and train a MLP
using the network structure file grng.top, which is shown below.
3
16 15 4
1 1
The network will have 3 layers including 16 inputs, 15 hidden units
in one hidden layer, and 4 outputs (one per class). In addition,
layers 2 and 3 connect to all previous layers. Training will stop
after 15 iterations, or when the classification error % reaches 4 % .
The final network weights will be stored in the file grng.wts.
c. Exit the DOS editor and observe the program running
d. Go to the "Examine Program Output" option and press <ret>
e. You can run this program on your own data, simply by editing the
parameter file in the "batch Run" option.