Go to the previous, next section, table of contents.
A structure that can be bound to an underlying network that implements a learning algorithm. The algorithm can be applied in either batch or online mode, according to <mode>. The learning rate is defined both through a default rate <lrate>, and for specific parameters, either for the ith parameter using <lrate:i> or for all (to be created) parameters whose name matches <id> using <lrate-map>. In addition, the learning rates can be adaptable, using <factor> and <factorminimum>.
Configuration
lrate=<value> sets default learning rate (0.01) adaptfactor=<> sets lr adaptation factor (0) factor=<v> sets lr factor lrate:<i>=<v> learning rate for parameter <i> lrate-map=<id>,<v> learning rate map factorminimum=<v> sets lr factor minimum (0.0001) mode=online|batch sets the training mode (online)Source Points
param:<i> the parameter <i> delta:<i> the delta for parameter <i>Trigger points
post_iter after each application post_iter_f after each application (front of queue)
Go to the previous, next section, table of contents. Document generated Mon Jun 16 02:19:05 GMT 2003