LPPool1d

class torch.nn.LPPool1d(norm_type, kernel_size, stride=None, ceil_mode=False)

Applies a 1D power-average pooling over an input signal composed of several input planes.

On each window, the function computed is:

$$f(X) = \sqrt[p]{\sum_{x \in X} x^{p}}$$

• At p = $\infty$, one gets Max Pooling over absolute values

• At p = 1, one gets Sum Pooling (which is proportional to Average Pooling)

Note

If the sum to the power of p is zero, the gradient of this function is not defined. This implementation will set the gradient to zero in this case.

Parameters:

• norm_type (int or float) – the power $p$ used to compute pooling. Must be non-zero; positive and negative infinity are supported.

• kernel_size (int) – the size of the window.

• stride (int or tuple of int, optional) – the stride of the window. Default: kernel_size.

• ceil_mode (bool, optional) – when True, use ceil instead of floor to compute the output shape. Default: False.

Note

When ceil_mode is True, sliding windows may go off-bounds if they start within the left padding or the input. Sliding windows that would start in the right padded region are ignored.

Shape:

• Input: $(N, C, L_{in})$ or $(C, L_{in})$.

• Output: $(N, C, L_{out})$ or $(C, L_{out})$, where

  $$L_{out} = \left\lfloor\frac{L_{in} - \text{kernel\_size}}{\text{stride}} + 1\right\rfloor$$

Examples::

>>> # power-2 pool of window of length 3, with stride 2.
>>> m = nn.LPPool1d(2, 3, stride=2)
>>> input = torch.randn(20, 16, 50)
>>> output = m(input)

forward(input)

Runs the forward pass.

Return type:

Tensor