tirank.Loss.gaussian_kernel

tirank.Loss.gaussian_kernel(a, b, sigma=1.0)[source]

Calculates the Gaussian (RBF) kernel similarity between two tensors.

Parameters:

  • a (torch.Tensor) – First input tensor (samples x features).

  • b (torch.Tensor) – Second input tensor (samples x features).

  • sigma (float, optional) – The sigma value (bandwidth) of the Gaussian kernel. Defaults to 1.0.

Returns:

The pairwise Gaussian kernel similarity matrix.

Return type:

torch.Tensor