Conv1d_cdopt

Conv1d_cdopt#

CLASS cdopt.nn.Conv1d_cdopt(in_channels, out_channels, kernel_size, stride=1, padding=0, dilation=1, groups=1, bias=True, padding_mode='zeros', device=None, dtype=None, manifold_class = euclidean_torch, penalty_param = 0,  weight_var_transfer = None, manifold_args = {})

Applies a 1D convolution over an input signal composed of several input planes. The basic introduction to convolution can be found at torch.nn.Conv1d.

This module supports TensorFloat32.

  • stride controls the stride for the cross-correlation, a single number or a one-element tuple.

  • padding controls the amount of padding applied to the input. It can be either a string {‘valid’, ‘same’} or a tuple of ints giving the amount of implicit padding applied on both sides.

  • dilation controls the spacing between the kernel points; also known as the à trous algorithm. It is harder to describe, but this link has a nice visualization of what dilation does.

  • groups controls the connections between inputs and outputs. in_channels and out_channels must both be divisible by groups. For example,

    • At groups=1, all inputs are convolved to all outputs.

    • At groups=2, the operation becomes equivalent to having two conv layers side by side, each seeing half the input channels and producing half the output channels, and both subsequently concatenated.

    • At groups= in_channels, each input channel is convolved with its own set of filters (of size \(\frac{\text{out\_channels}}{\text{in\_channels}}\)).

Parameters:#

  • in_channels (int) – Number of channels in the input image

  • out_channels (int) – Number of channels produced by the convolution

  • kernel_size (int or tuple) – Size of the convolving kernel

  • stride (int or tuple, optional) – Stride of the convolution. Default: 1

  • padding (int, tuple or str, optional) – Padding added to both sides of the input. Default: 0

  • padding_mode (string*,* optional) – 'zeros', 'reflect', 'replicate' or 'circular'. Default: 'zeros'

  • dilation (int or tuple, optional) – Spacing between kernel elements. Default: 1

  • groups (int, optional) – Number of blocked connections from input channels to output channels. Default: 1

  • bias (bool, optional) – If True, adds a learnable bias to the output. Default: True

  • manifold_class – The manifold class for the weight matrix. Default: cdopt.manifold_torch.euclidean_torch

  • penalty_param – The penalty parameter for the quadratic penalty terms in constraint dissolving function

  • manifold_args - The additional key-word arguments that helps to define the manifold constraints.

  • weight_var_transfer (callable) – The function that transfer the weights (3D-tensor) to the shape of the variables of the manifold.

    • The weight_var_transfer is called by
      weight_to_var, var_to_weight, var_shape = weight_var_transfer( tensor_shape )

    • The inputs of weight_var_transfer should be the size of the weights. As for the outputs, weight_to_var is the callable function that transfer the weights to the variables of the manifold. var_to_weight is the callable function that transfers the variables of the manifold to the weights. var_shape is a tuple of ints that refers to the shape of the variables of the manifolds.

    • Default:

      • var_shape = (torch.prod(torch.tensor(tensor_shape[:-1])), torch.tensor(tensor_shape[-1]))

      • weight_to_var = lambda X_tensor: torch.reshape(X_tensor, var_shape)

      • var_to_weight = lambda X_var: torch.reshape(X_var, tensor_shape)

Shapes:#

  • Input: \((N, C_{in}, L_{in})\) or \((C_{in}, L_{in})\).

  • Output: \((N, C_{out}, L_{out})\) or \((C_{out}, L_{out})\), where \(L_{out} = \left\lfloor\frac{L_{in} + 2 \times \text{padding} - \text{dilation} \times (\text{kernel\_size} - 1) - 1}{\text{stride}} + 1\right\rfloor\).

Attributes:#

  • manifold (cdopt manifold class) – the manifold that defines the constraints. The shape of the variables in manifold is set as var_shape.

  • weight (Tensor) – the learnable weights of the module of shape \((\mathrm{out\_channels}, \frac{\text{in\_channels}}{\text{groups}}, \mathrm{kernel\_size})\). The values are initialized from var_to_weight(manifold.Init_point(weight_to_var(Xinit))), where \(\mathrm{Xinit}\sim \mathcal{U}(-\sqrt{k}, \sqrt{k})\) where \(k = \frac{\mathrm{groups}}{C_\text{in} * \mathrm{kernel\_size}}\).

  • bias (Tensor) – the learnable bias of the module of shape (out_channels). If bias is True, then the values of these weights are sampled from \(\mathcal{U}(-\sqrt{k}, \sqrt{k})\) where \(k = \frac{\mathrm{groups}}{C_\text{in} * \text{kernel\_size}}\).

  • quad_penalty (callable) – the function that returns the quadratic penalty terms of the weights. Its return value equals to \(||\mathrm{manifold.C}(\mathrm{weight})||^2\).

Example:#

my_layer = cdopt.nn.Conv1d_cdopt(16, 33, 3, stride=2, manifold_class= cdopt.manifold_torch.stiefel_torch)
input = torch.randn(20, 16, 50)
output = my_layer(input)
print(output.size())
Contents