Wrong equations for the two neuron toy system

[ft-major]

Hi,
I was studying your very interesting paper and looking into the code you published (thanks for that) there are some things that I don't understand or are different from what is written in the paper.

For example, the first thing that I noticed is that the equation for simulating neuron B in testSystem.m doesn't have the noise multiplied by t2. Now it is
for t = 2:MAX_TIME
x(t) = x(t-1) + ((-x(t-1) + sigmf(w11*x(t-1) + w12*y(t-1) - b1, [1 0]))/t1 + normrnd(0, NOISE_SIGMA))*dt;
y(t) = y(t-1) + ((-y(t-1) + sigmf(w21*x(t-1) + w22*y(t-1) - b2, [1 0]))/t2 + normrnd(0, NOISE_SIGMA))*dt;
end
while it should be
for t = 2:MAX_TIME
x(t) = x(t-1) + ((-x(t-1) + sigmf(w11*x(t-1) + w12*y(t-1) - b1, [1 0]))/t1 + normrnd(0, NOISE_SIGMA))*dt;
y(t) = y(t-1) + ((-y(t-1) + sigmf(w21*x(t-1) + w22*y(t-1) - b2, [1 0]) + normrnd(0, NOISE_SIGMA))/t2)*dt;
end
Another thing that is not clear to me is why, as you specified at the begining of testSystem.m script, the results are seed dependet. It is normal? For example changing the seed transform the phase of the dominant eigenvalue from this

to this

and the loop identification algorithm gives completely different results.

Thanks in advance for the answer.

P.s. is this a good place to ask more questions about the mathematical part, or you prefer other communication channels?