Re: [Pcsim-users] plot

[Pecevski D. <de...@ig...>]

Mathieu Dubois wrote:
> I want to build (and plot) the same LSM than in [1] (very similar to the 
> one given in the Circuit-Tool tutorial).
> I have created the liquid neurons as a spatial population with 2 families:
> liquid_nrn_popul = SpatialFamilyPopulation( net, [ exc_nrn_factory, 
> inh_nrn_factory ], RatioBasedFamilies( EXC_INH_Ratio  ), 
> CuboidIntegerGrid3D(LIQUID_SIZE, LIQUID_ORIGIN) );
> so I can make the connections inside the liquid with a 
> ConnectionsProjection and draw those synapses with the help you gave me.
>
> But for connecting (for instance) the input neurons to the liquid I use 
> net.connect (as in the examples) so I cannot use a ConnectionsProjection 
> (each input neuron is connected to the three liquid neurons at the same 
> altitude).
>   
It's not clear to me why you want  to use a ConnectionsProjection?
If I understood correctly, the connection map from input to the liquid is
predefined, you already know which neurons to connect in the plot, so you
probably don't need this to get the presynaptic and postsynaptic neuron 
pairs.

If you want to get the parameters of the synapses you can achieve this with
the synapses IDs.

When you execute

syn_id = net.connect( pre_nrn_id, post_nrn_id, StaticSpikingSynapse() )

for example, then the output of net.connect is the id of the synapse 
created.
Now you can get this synapse with

net.object(syn_id)

regards,
Dejan
> I think that I can define a new sub-population in the liquid with the 
> function subset (and therefore  use a ConnectionsProjection) but I 
> cannot figure out how to do that with the online help.
>   
> So my question is: How to define a subpopulation?
>   
You can define a new population by giving a list of IDs of the simobjects:

---------------

exz_nrn_popul = SimObjectPopulation(net, lifmodel, int( nNeurons *  Frac_EXC ) );
inh_nrn_popul = SimObjectPopulation(net, lifmodel, nNeurons - exz_nrn_popul.size() );

all_nrn_popul = SimObjectPopulation(net, list(exz_nrn_popul.idVector()) + list(inh_nrn_popul.idVector()));

print "Created", exz_nrn_popul.size(), "exz and", inh_nrn_popul.size(), "inh neurons";
------------
See example : http://www.lsm.tugraz.at/pcsim/examples/cuba_model_populations.py

You can extract the ids of the neurons in the liquid that the input neurons connect to, with 

id = liquid_popul.getIdAt( 3,4,5 ).packed()

make a python list of these ids 

and then call

new_popul = SimObjectPopulation( net, id_list )

And then you can create the wanted projections between populations.

I hope that's helpful.

regards, 
Dejan

> Thanks in advance,
> Mathieu
>
> [1] "Imitation Learning with spiking networks and real-world devices", 
> H. Burgsteiner, Engineering Applications of AI, 19, 741-752, 2006
>
>