Matthew Aburn

As a workaround in the meantime, `std::array` with the experimental nvcc flag `--expt-relaxed-constexpr` works for me. (interested to know if there are cases where this breaks.)

Ok, I've contributed test coverage for stdex::extents in pull request #38 This identifies a bug in one constructor which is fixed in pull request #36

I agree it would be useful to include support for jump processes. Will consider how to implement this in future, and I'd welcome your suggestions on the interface.

Thanks for the suggestion. That is a nice approach for DDE (using linear interpolation of history together with an existing integrator) but I think that code design can not be...

You are right, examples in the README would be good. Love the `@interact`, works very well for demoing stochastic processes. If I can find a way to handle .ipynb files...

For a 2-dimensional system nsim expects the state y to be an array of shape (2,) Your function f() returns an array of shape (2,1). If you remove the extra...

The fact that you got an error message from numpy instead of a more helpful message means that nsim could use some better input validation. So this is a bug...

Yes, superconvergence is right. You pre-generate the noise, then pass it in to `itoSRI2()` or `itoEuler()`. You can use the function `sdeint.deltaW(n, m, h)` to pre-generate the noise increments. (It...

Yes, in that case the G function should return a constant dxd diagonal array. Very roughly speaking, the entries correspond to std, not variance. More precisely, if function f were...

Thanks, I think that is a useful suggestion, so that we can provide an interface as close to `odeint` as possible. I'll implement this when I can get some time....