Document how to record population averages in a memory efficient way

[paulbrodersen]

Hi, following our discussion here, I added a small section to the documentation that outlines how to continuously record population averages in a memory efficient way (so I can find the information next time I need it).

The section is based on the example given by @mstimberg here.

group = NeuronGroup(..., 'dv/dt = ...', ...)

# Dummy group to store the average membrane potential at every time step
vm_container = NeuronGroup(1, 'average_vm : volt')
# Synapses averaging the membrane potential of all neurons
vm_averager = Synapses(group, average_vm, 'average_vm_post = v_pre/N_pre : volt (summed)')
vm_averager.connect()
# Monitor recording the average membrane potential
vm_mon = StateMonitor(vm_container, 'average_vm', record=0)

I made the following changes:

1. I indicated that the units of the variable and the variable average need to match: group = NeuronGroup(..., 'dv/dt = ... : volt', ...)
2. I corrected a typo: vm_averager = Synapses(group, vm_container, ...)
3. I set record=True to match the examples in the same section.

[mstimberg]

Looks great to me, many thanks for this contribution! I changed a minor detail: if I am not mistaken, the memory usage in your example would be even worse since each value is a double, i.e. takes 8 bytes. So recording 1000 × 100 × 10000 (neurons × duration × timesteps per second) values would take up ~ 8 Gigabytes 😬

[paulbrodersen]

Apparently, I hadn't had enough coffee this morning, because I thought a float64 would only need 8 bits not 8 bytes...
Thanks for the correction!