In the neuron method that is called by the synapse to obtain postsynaptic trace values, recompute_internal_variables() is called twice, updating all neuron state variables. Typically, only one or a few of these are actually necessary to be updated. A large runtime performance gain can be gained. To implement this, we probably need to do a recursive dependency analysis of all the state variables that the requested trace variable depends on. Could this mean that the neuron could end up in an inconsistent state, with some variables at timepoint a and others at timepoint b?
In the neuron method that is called by the synapse to obtain postsynaptic trace values,
recompute_internal_variables()is called twice, updating all neuron state variables. Typically, only one or a few of these are actually necessary to be updated. A large runtime performance gain can be gained. To implement this, we probably need to do a recursive dependency analysis of all the state variables that the requested trace variable depends on. Could this mean that the neuron could end up in an inconsistent state, with some variables at timepoint a and others at timepoint b?