Control vector U with unkown function[BUG] about pysindy

[briandesilva]

See this section of the documentation. The following factors contribute to the issue:

1. The default numerical integrator used by SINDy.simulate is scipy.integrate.odeint, an adaptive time-stepping routine. Since odeint is adaptive, it needs to be able to evaluate functions at various points in time, meaning control inputs need to be given as a callable function.
2. Sometimes you don't have a function for u, but you do have a time series. In that case, SINDy will form a spline interpolation of the time series and use that instead. However, since it's an interpolating function, if you try to evaluate it at a time outside of the original time series, it will complain.

The issue is either that you're trying to simulate for times when the control inputs aren't known or that the numerical integrator, odeint, is attempting to evaluate the interpolating function for u outside the domain where it's defined. You could either decrease the span of time over which you simulate or use a different integrator (scipy.integrate has some other options). If you decide to do the latter, you will probably need to adjust the call signature of the integrator to match that which SINDy.simulate expects: integrator(rhs, x0, t, **kwargs).

[msis]

This is what I used to get too, see #94 .
I think it's related to the t you are passing, not the u.

[briandesilva]

This is what I used to get too, see #94 .
I think it's related to the t you are passing, not the u.

That could also be the problem. I will create a new release including the fix this afternoon.

[moujrid]

thank you for all you replies.

I tried to shift the time to start at 0 and also tried different time spans smaller than the actual time span of my time serie but the error keeps showing up

[briandesilva]

@moujrid, I just created a new release with the bug fix I alluded to earlier. See if you still encounter the error after updating to the latest version.