Hebbian learning across two ensembles

General Discussion
1

Hi,

I have been trying to apply hebbian learning across two ensembles but the system shows a period of synchrony and desynchrony between the ensembles. I want to have synchrony behaviour in the system after learning. Here is the nengo code, is there anything which I am doing wrong?

import nengo
import numpy as np

model = nengo.Network('stngpeloop')
synapse=0.1
lam=3
wstn2gpe = 3
wstn2stn = 2
    
def dynamics(x):
    return (synapse*(-x[0] + wstn2stn*np.tanh(lam * x[0]) - wstn2gpe*x[1]) + x[0]
              , synapse*(-x[1] + wstn2gpe*np.tanh(lam * x[0])) + x[1])    
with model:
    stim = nengo.Node([0,0])
    stngpe = nengo.Ensemble(200,2,radius = 1)
    stngpe2 = nengo.Ensemble(200,2,radius = 1)
    nengo.Connection(stim,stngpe)
    nengo.Connection(stngpe,stngpe,function = dynamics)
    
    nengo.Connection(stim,stngpe2)
    nengo.Connection(stngpe2,stngpe2,function = dynamics)
    
    nengo.Connection(stngpe[0],stngpe2[0],function = lambda x: 0,learning_rule_type = nengo.Oja(learning_rate = 1e-15,beta=0,pre_tau = 0.005),
                    solver=nengo.solvers.LstsqL2(weights = True))
                    
    nengo.Connection(stngpe2[0],stngpe[0],function = lambda x: 0,learning_rule_type = nengo.Oja(learning_rate = 1e-15,beta=0,pre_tau = 0.005),
                    solver=nengo.solvers.LstsqL2(weights = True))
2

There’s no defined way to create synchrony (I’m assuming you mean synchrony between neurons and their spiking patterns) between Nengo Ensembles (let alone learn synchrony), because we don’t (and I would argue that no one does) have a really good understanding of what synchrony does computationally. There is one idea of how to implement it, proposed by Terry in our “Modelling Ideas” repository.

Does this somewhat answer your question?

Also, sorry it took us so long to respond to this. Summer is a weird time for our lab in terms of vacations and conferences.