I download the rbs.master and tested it. There isn't a spike train comparison however, it seems to run. However, the standard version of the stateMachineClass file differs slightly from the RBS version. There are some minor things but two constants differ.
rbs NEAL
self.STATE_TO_ONE_WEIGHT = .015 0.01
self.ONE_HALF_ON_ONE_WEIGHT = 0.01 0.08

These are only used in stateMachineClass.stateTurnsOnOneNeuron and stateMachineClass.oneNeuronHalfTurnsOnOneNeuron.

As far as I can tell these are only used in the rbs in the tests. It looks like they still work, but you should probably check. If that works we can both use the same variables, which is as it should be. If not, we'll have to introduce two (or one) new variable.

Add ability for Rules to interact with Neural Timer.

AC:

• Rules can start a unique instance of a timer depending on variables.
• Rules can stop a unique instance of a timer depending on variables.
• Condition exits to evaluate an instance of a timer.

Instead of calling the architecture via a python file, we should be able to write a text file that you pass to the architecture python root.

associationBuilder.useBases has an incorrect variable name. It should be baseFileName not bases. Fix that and the plethora of other poorly named variables in associationBuilder

Currently the "root" of the system is rbsBuilder, and this isn't accurate. The idea is that the cognitive architecture contains everything including the rbs and associative memory. So, the root of the system should be cogArchBuilder or nArcBuilder or something like that.

Modify the interaction between the associative memory and rule based system so that we can access properties, bases, and operations in the rules. This should be the yes answers for the Quillian task.

Instead of having FSAs and NEAL in different parts of the architecture, we should use just one, and instantiate them just once. In particular, NEAL resolves issues about multiple synapse connection list from populationA to populationB on SpiNNaker. Also, want to have the current (or near current) stateMachine class.

Currently it is not clear how the cog arc builds the neural networks