How define Ensembles with Non-Zero value?

#1

Hi,
I would like to represent a Quaternion (a 4D variable) with a 4-dimensional ensemble and use it in the subsequent calculations.
The variable which has been represented with this ensemble (A in my codes) is denominator in my calculations and because its value is zero in the first few iterations, a “Zero Division Error” is returned.
How the ensemble’s parameters could be adjusted to return a non-zero variable? This error could be eliminated if the ensemble’s parameters are properly adjusted, but which parameters are they?

A [[ 0. 0. 0. 0. ]
[ 0. 0. 0. 0. ]
[-0.00615827 0.00846977 0.00803852 0.01110903]
…
[-0.357 0.491 0.466 0.644 ]
[-0.357 0.491 0.466 0.644 ]
[-0.357 0.491 0.466 0.644 ]]

The codes are :

import nengo
    from nengo.processes import Piecewise
    import matplotlib.pyplot as plt
    from pyquaternion import Quaternion


    q1 = Quaternion(-0.357 ,0.491,0.466, 0.644)
    v1=q1.elements

    model = nengo.Network()
    with model:
	ndA = {0:[0,0,0,0], 0.2:v1}

	inputA = nengo.Node(Piecewise(ndA))

        inputA = nengo.Node(v1)

        A = nengo.Ensemble(400, dimensions=4,neuron_type=nengo.Direct())


        B =nengo.Ensemble(400, dimensions=4, neuron_type=nengo.Direct())

        nengo.Connection(inputA, A)

        def inverse(C):
            
            q = Quaternion(a=C[0], b=C[1], c=C[2], d=C[3])
            q_inv=q.inverse
            inv= q_inv.elements

            return inv

        nengo.Connection(A,B, function=inverse)


        A_probe = nengo.Probe(A, synapse=0.01)

        B_probe = nengo.Probe(B, synapse=0.01)


sim_time = 1
sim = nengo.Simulator(model)
sim.run(sim_time)

plt.figure()
plt.plot(sim.trange(), sim.data[A_probe])
print("A", sim.data[A_probe])

plt.show()
plt.figure()
plt.plot(sim.trange(), sim.data[B_probe])
print("BB", sim.data[B_probe])
plt.show()
#2

You can write whatever code you want inside the inverse function, so I would just add an explicit check in there to guard against 0 denominators. Like if q == 0: q_inv = 0 or something like that, however you want to handle that for your application.