Fills the input Tensor with a (semi) orthogonal matrix, as
described in Exact solutions to the nonlinear dynamics of learning in deep linear neural networks - Saxe, A. et al. (2013). The input tensor must have
at least 2 dimensions, and for tensors with more than 2 dimensions the
trailing dimensions are flattened.
Examples
if (torch_is_installed()) {
w <- torch_empty(3, 5)
nn_init_orthogonal_(w)
}
#> torch_tensor
#> 0.0822 -0.7016 -0.5799 0.2274 0.3361
#> 0.9028 0.2638 -0.2770 -0.1959 -0.0155
#> -0.3268 0.1279 -0.6633 -0.3514 -0.5598
#> [ CPUFloatType{3,5} ]