neural_ode_mlp#

diffeqzoo.ivps.neural_ode_mlp(*, initial_values=(0.0,), time_span=(0.0, 1.0), scale=1.0, layer_sizes=(2, 20, 1), seed=1234)[source]#

Construct an IVP with a neural ODE vector field.

The vector field is given by a neural network.

The neural network is a multi-layer perceptron with tanh activation functions.

We implement the dynamics used in the “implicit-layers” tutorial: http://implicit-layers-tutorial.org/neural_odes/.

Note

The neural network is not trained in this function. This function only constructs the IVP.