From 41fa43401c9dcc77b2c41450d2be118ea6fa7fe2 Mon Sep 17 00:00:00 2001 From: Sathvik Bhagavan Date: Thu, 4 Apr 2024 04:17:51 +0000 Subject: [PATCH] docs: add reference for the bloch equations in the complex example --- docs/src/examples/complex.md | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/docs/src/examples/complex.md b/docs/src/examples/complex.md index a44e31e79f..a11a4d4d0f 100644 --- a/docs/src/examples/complex.md +++ b/docs/src/examples/complex.md @@ -1,6 +1,6 @@ # Complex Equations with PINNs -NeuralPDE supports training PINNs with complex differential equations. This example will demonstrate how to use it for [`NNODE`](@ref). Let us consider a system of [bloch equations](https://en.wikipedia.org/wiki/Bloch_equations). Note [`QuadratureTraining`](@ref) cannot be used with complex equations due to current limitations of computing quadratures. +NeuralPDE supports training PINNs with complex differential equations. This example will demonstrate how to use it for [`NNODE`](@ref). Let us consider a system of [bloch equations](https://en.wikipedia.org/wiki/Bloch_equations) [^1]. Note [`QuadratureTraining`](@ref) cannot be used with complex equations due to current limitations of computing quadratures. As the input to this neural network is time which is real, we need to initialize the parameters of the neural network with complex values for it to output and train with complex values. @@ -95,3 +95,5 @@ plot!(ground_truth.t, imag.(reduce(hcat, ground_truth.u)[4, :])) ``` We can see it is able to learn the real parts of `u1`, `u2` and imaginary parts of `u3`, `u4`. + +[^1]: https://steck.us/alkalidata/