Complex Differentiation

Complex numbers add some difficulty to the idea of a "gradient". To talk about gradient(f, x) here we need to talk a bit more about f.

If f returns a real number, things are fairly straightforward. For $c = x + yi$ and $z = f(c)$, we can define the adjoint $\bar c = \frac{\partial z}{\partial x} + \frac{\partial z}{\partial y}i = \bar x + \bar y i$ (note that $\bar c$ means gradient, and $c'$ means conjugate). It's exactly as if the complex number were just a pair of reals (re, im). This works out of the box.

julia> using Zygote

julia> gradient(c -> abs2(c), 1+2im)
(2.0 + 4.0im,)

However, while this is a very pragmatic definition that works great for gradient descent, it's not quite aligned with the mathematical notion of the derivative: i.e. $f(c + \epsilon) \approx f(c) + \bar c \epsilon$. In general, such a $\bar c$ is not possible for complex numbers except when f is holomorphic (or analytic). Roughly speaking this means that the function is defined over c as if it were a normal real number, without exploiting its complex structure – it can't use real, imag, conj, or anything that depends on these like abs2 (abs2(x) = x*x'). (This constraint also means there's no overlap with the Real case above; holomorphic functions always return complex numbers for complex input.) But most "normal" numerical functions – exp, log, anything that can be represented by a Taylor series – are fine.

Fortunately it's also possible to get these derivatives; they are the conjugate of the gradients for the real part.

julia> gradient(x -> real(log(x)), 1+2im)[1] |> conj
0.2 - 0.4im

We can check that this function is holomorphic – and thus that the gradient we got out is sensible – by checking the Cauchy-Riemann equations. In other words this should give the same answer:

julia> -im*gradient(x -> imag(log(x)), 1+2im)[1] |> conj
0.2 - 0.4im

Notice that this fails in a non-holomorphic case, f(x) = log(x'):

julia> gradient(x -> real(log(x')), 1+2im)[1] |> conj
0.2 - 0.4im

julia> -im*gradient(x -> imag(log(x')), 1+2im)[1] |> conj
-0.2 + 0.4im

In cases like these, all bets are off. The gradient can only be described with more information; either a 2x2 Jacobian (a generalisation of the Real case, where the second column is now non-zero), or by the two Wirtinger derivatives (a generalisation of the holomorphic case, where $\frac{∂ f}{∂ z'}$ is now non-zero). To get these efficiently, as we would a Jacobian, we can just call the backpropagators twice.

function jacobi(f, x)
  y, back = Zygote.pullback(f, x)
  back(1)[1], back(im)[1]
end

function wirtinger(f, x)
  du, dv = jacobi(f, x)
  (du' + im*dv')/2, (du + im*dv)/2
end
julia> wirtinger(x -> 3x^2 + 2x + 1, 1+2im)
(8.0 + 12.0im, 0.0 + 0.0im)

julia> wirtinger(x -> abs2(x), 1+2im)
(1.0 - 2.0im, 1.0 + 2.0im)