Fully connected network used by pinnr

A multilayer perceptron with Xavier-normal initial weights and zero biases, the standard choice for PINNs.

Usage

pinn_mlp(n_in, hidden = c(32, 32, 32), n_out = 1, activation = "tanh")

Arguments

n_in

Number of inputs.

hidden

Integer vector of hidden layer widths.

n_out

Number of outputs.

activation

"tanh" (default), "sin", "silu", "gelu" or "softplus". Smooth activations are needed because the loss involves derivatives of the network; avoid ReLU.

Value

A torch::nn_module.

Examples

net <- pinn_mlp(2, c(16, 16), 1)
net(torch::torch_randn(5, 2))
#> torch_tensor
#> -0.6970
#> -0.2253
#> -0.0910
#>  0.6885
#> -0.5907
#> [ CPUFloatType{5,1} ][ grad_fn = <AddmmBackward0> ]