mlx.nn.Bilinear

Contents

mlx.nn.Bilinear#

class Bilinear(input1_dims: int, input2_dims: int, output_dims: int, bias: bool = True)#

Applies a bilinear transformation to the inputs.

Concretely:

\[y_i = x_1^\top W_i x_2 + b_i\]

where: \(W\) has shape [output_dims, input1_dims, input2_dims], \(b\) has shape [output_dims ], and \(i\) indexes the output dimension.

The values are initialized from the uniform distribution \(\mathcal{U}(-{k}, {k})\), where \(k = \frac{1}{\sqrt{D_1}}\) and \(D_1\) is input1_dims.

Parameters:

  • input1_dims (int) – The dimensionality of the input1 features

  • input2_dims (int) – The dimensionality of the input2 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.

Methods

Contents