Julia vs NumPy broadcasting

[stevengj]

@baggepinnen

any numerical error or difference in initial condition etc. should be forgotten exponentially

Only in an absolute-error sense — the relative error still accumulates if you compute e^{Ak} by repeated multiplication.