class Linear(input_dims: int, output_dims: int, bias: bool = True)#

Applies an affine transformation to the input.


\[y = x W^\top + b\]

where: where \(W\) has shape [output_dims, input_dims] and \(b\) has shape [output_dims].

The values are initialized from the uniform distribution \(\mathcal{U}(-{k}, {k})\), where \(k = \frac{1}{\sqrt{D_i}}\) and \(D_i\) is equal to input_dims.

  • input_dims (int) – The dimensionality of the input features

  • output_dims (int) – The dimensionality of the output features

  • bias (bool, optional) – If set to False then the layer will not use a bias. Default is True.