Overview of havBpNet-J - Restoring a Network

havBpNet:J

Intro    Network Classes Overview    Data Classes Overview    Packages

havBpNet:J Home Page

Restoring a Network

Restoring havBpNet nodelists, layers and nodes is accomplished using alternate constructors defined for the havBpNodeList, havBpLayer and havBpNode classes. Use of these alternate constructors is illustrated in the following code excerpt which restores our xortrain.java example network:.


 net_file = new FileInputStream(netfile);           // open the SAVE file

 biasValue = havSys.readDouble(net_file);           // first restore the private information
 havBpNodeList net = new havBpNodeList(net_file);   // followed by the nodelist and
 havBpLayer bias   = new havBpLayer(net, net_file); // then restore the network layers
 havBpLayer in     = new havBpLayer(net, net_file); // using the same order in which they
 havBpLayer hid    = new havBpLayer(net, net_file); // were saved
 havBpLayer out    = new havBpLayer(net, net_file);
 net_file.close();

It is the user’s responsibility to open the save file and to ensure that the file pointer is appropriately positioned for the restore operation. Correct file pointer positioning can most easily be ensured by performing the restore in the same order that was used during the network save.

Also notice the explicit read required to handle the application-specific information that we saved in the SAVE file.

Notice that there is no need to perform explicit Connect(...) operations on the restored network. This is accomplished automatically as the network’s layers are restored.

Layer creation initializes the layer with training enabled. If the network is being restored for consultation and if there will be Train() messages sent to any of the layers for error determination, it is the user’s responsibility, after restoring a network, to ensure that weight adjustment is ennabled or disabled as needed.

Overview || Create || Train || Save || Restore || Consult || Data Scaling

Intro    Network Classes Overview    Data Classes Overview    Packages