I am not aware of an option for adjusting the learning rate in the HPNEURAL procedure but you can adjust the learning rate in the Neural Network node in SAS Enterprise Miner. To do so, drag a Neural Network node onto your SAS Enterprise Miner diagram and then click on the ... to the right of Optimization properties in the Neural Network node properties sheet. From the dialog which appears, change the Training Technique property from Default to BackProp, QPROP, or RPROP and you will then be able to modify the Learn property:
Learn - Specifies the learning rate for BackProp or the initial learning rate for QPROP and RPROP optimization techniques.
When you specify RPROP which uses a separate learning rate for each weight and adapts all the learning rates during training, the dialog allows customization of some additional options including
Accelerate - Specifies the rate of increase of learning for the RPROP optimization technique.
Decelerate - Specifies the rate of decrease of learning for the RPROP optimization technique.
Maximum Learning - Specifies the maximum learning rate for RPROP.
Minimum Learning - Specifies the minimum learning rate for RPROP.
When you specify QPROP (for QuickProp) which is a Newton-like method thatuses a diagonal approximation to the Hessian matrix, the dialog allows customization of the some additional options including
Maximum Momentum - Specifies the maximum momentum rate for QPROP.
Tilt - Specifies the tilt value for QPROP.
The direct use of the NEURAL procedure which is used by the Neural Network node is not supported, however, so you would need to build this type of model using the SAS Enterprise Miner interface.
Hope this helps!
Doug
