Brian Ward

This means one of the network calls on the board failed. If you have debug enabled in your config, the output of the program should log which one is failing

I believe I suggested to the contrary on slack recently, but I should note that the user defining functions called things like `add`, `subtract`, etc right now does _not_ allow...

I ported @bob-carpenter's code above to use the template metaprograms required in this branch: https://github.com/stan-dev/math/tree/feature/2704-schur-decomposition The mix tests, including the ones originally commented out, all pass. Unfortunately, I tried to...

I can't convince wolfram alpha to output the complex Schur decomposition, only the real one

I've been using the example from scipy: https://docs.scipy.org/doc/scipy/reference/generated/scipy.linalg.schur.html `[[0, 2, 2], [0, 1, 2], [1, 0, 1]]` On WA just now I tried `[[0, 2, 2], [0, 1i, 2], [1,...

The decomposition returned by Eigen does satisfy `A = U T U*`, so maybe the unit test is just that? ```c++ Eigen::MatrixXcd A(3, 3); A

That's what I get for not making it down into the "Notes" section of Wikipedia. Thanks

Found some more signatures we would probably like to support but that don't compile: - [ ] `subtract(complex, complex_matrix) => complex_matrix` https://gist.github.com/WardBrian/0a9cf4651a844165e325fbb208fde7dd - [ ] `add(complex, complex_matrix) => complex_matrix` https://gist.github.com/WardBrian/67f3f279d19c386998caaafac36c0cf9

> Does the other order, `subtract(complex_matrix, complex)`, work? No, the error is essentially the same

Does adding specializations for the analytic derivatives of FFT/iFFT not fix the test issues?