qfeval_functions.functions.nanmean

nanmean(x, dim=(), keepdim=False)[source]

Compute the arithmetic mean along specified dimensions, ignoring NaN values.

This function calculates the mean (average) of tensor elements along the specified dimension(s), excluding NaN values from the computation. NaN values are treated as missing data and do not contribute to either the sum or the count used in the mean calculation.

The NaN-aware mean is computed as:

\[\text{nanmean}(X) = \frac{\sum_{i \text{ valid}} X_i}{N_{\text{valid}}}\]

where the sum is over valid (non-NaN) values and \(N_{\text{valid}}\) is the number of valid values.

Parameters:

x (Tensor) – The input tensor containing values.
dim (Union[int, Tuple[int, ...]]) – The dimension(s) along which to compute the mean. If not specified (default is empty tuple), the mean is computed over all dimensions.
keepdim (bool) – Whether the output tensor has dim retained or not. Default is False.

Returns:

The mean values computed only over valid (non-NaN) values. The shape depends on the input dimensions, dim, and keepdim parameters.

Return type:

Tensor

Example

>>> # Simple mean with NaN values
>>> x = torch.tensor([1.0, 2.0, nan, 4.0, 5.0])
>>> QF.nanmean(x)
tensor(3.)

>>> # 2D tensor with NaN values
>>> x = torch.tensor([[1.0, nan, 3.0],
...                   [4.0, 5.0, nan]])
>>> QF.nanmean(x, dim=1)
tensor([2.0000, 4.5000])

>>> # All dimensions
>>> x = torch.tensor([[1.0, 2.0],
...                   [nan, 4.0]])
>>> QF.nanmean(x)
tensor(2.3333)

>>> # With keepdim
>>> x = torch.tensor([[1.0, nan, 3.0],
...                   [4.0, 5.0, nan]])
>>> QF.nanmean(x, dim=1, keepdim=True)
tensor([[2.0000],
        [4.5000]])

>>> # All NaN slice
>>> x = torch.tensor([[1.0, 2.0],
...                   [nan, nan]])
>>> QF.nanmean(x, dim=1)
tensor([1.5000,    nan])