williamgilpin / dysts Goto Github PK

The second example in the README renders a blank visualization:

from dysts.flows import Lorenz

model = Lorenz()
model.gamma = 1
model.ic = [0, 0, 0.2]
sol = model.make_trajectory(1000)
plt.plot(sol[:, 0], sol[:, 1])

Results in:

Issue seems to be with the model.ic settings.

I am collecting a list of systems that I would like to implement. Please feel free to add any additional suggestions here.

+ Lid driven cavity flow
+ Interior squirmer model

Edit: both systems have now been added

Hi Dr.Gilpin,

Thank you for the amazing work and benchmark. I was wondering what the module named 'degas' is. I can't seem to find it on pypi or conda forge. Could you please advise?

Best,
Keun

Hi, I've been looking at the implementations of the Sakarya and LiuChen systems and I have a couple of questions. It's my understanding that, since the two systems are similar, the implementation is based on the same equations, simply changing the free parameters. Although looking at the equations in "A novel four-wing strange attractor born in bistability" the second equation reads as

while in the code it's implemented as ydot = -b * y - p * x + q * x * z. Should this be ydot = b * yz - p * y + q * x * z (also accounting for the LiuChen and adjusting the parameters) or there is something I am overlooking?

Also the third equation

is implemented as zdot = c * z - r * x * y. is there a reason to not include the +b?

Wouldn't be more straight forward to implement the two systems as separate to increase readability?

Sorry in advance if I missing something obvious, and thanks for this work I've been really enjoying the repo!

Wow, awesome job with this! I'm going to add an example of using these datasets with some of the PySINDy repository's new functionality -- will update you on progress.

Also, you may be interested in adding the chaotic "atmospheric oscillator" from Tuwankotta, J. M. (2003). Widely separated frequencies in coupled oscillators with energy-preserving quadratic nonlinearity. Physica D: Nonlinear Phenomena, 182(1-2), 125-149. Our group also uses a simple version of this system in Kaptanoglu, A. A., Callaham, J. L., Aravkin, A., Hansen, C. J., & Brunton, S. L. (2021). Promoting global stability in data-driven models of quadratic nonlinear dynamics. Physical Review Fluids, 6(9), 094401.

Reservoir Computing appears to be the SotA in many tasks involving dynamical systems, it would be interesting to have it among the models.

Hi Will,

Hope you're doing well. We are working on a PySINDy + dysts project and are wondering about the precise way the 'dt' parameter is calculated. It appears that the 'period' parameter is the largest timescale in the system, and 'dt' maybe the smallest timescale in the system, although in the paper you refer to instead as "optimal integration timestep", so I'm not sure. For instance, for the AtmosphericRegime my understanding from the Tuwankotta paper is that the fast timescale = 0.01 but for the dysts database dt = 0.01773980357826076. If it is not the smallest timescale, is it close? Is there a good way to calculate the true smallest timescale? Any clarification would be appreciated.

Best,
Alan

I'm sure there are reasons why the separate JSON file is convenient, but I found it made it hard to understand exactly what dynamical systems are being solved without installing the Python package.