Pass empty parameter to LearningRuleType

Backends
1

Is there any way of passing an empty parameter to a LearningRuleType? This would be useful, for example, if trying to run a model in a script taking command-line parameters. If no command-line parameter were specified (for ex. the PES learning rate) I would like the learning rule to use the default for its parameter:

NumberParam(“learning_rate”, low=0, readonly=True, default=1e-4)

I would then be able to instantiate the object in my script in a way similar to this:

conn.learning_rule_type = PES(learning_rate=empty())

None does not work with non-optional parameters and DefaultParameterType also is not accepted.

2

Most (if not all) Nengo objects can be instantiated with default values using the Default class from nengo.params. Here’s some example code specific to your question:

from nengo.params import Default

conn.learning_rule_type = PES(learning_rate=Default)