Graph Augmentations

hivegraph.augmentations.drop_feature(x: Tensor, drop_prob: float, dtype: dtype = torch.float32) → Tensor[source]

Implementation of Masking Node Features (MF)

Randomly masks a fraction of dimensions with zeros in node features.

Ref: Section 3.2.2 of https://arxiv.org/abs/2006.04131v2

Parameters:

  • x (torch.Tensor) – Node features.

  • drop_prob (float) – Probability of dropping a feature.

  • dtype (torch.dtype) – Data type of the tensor.

Returns:

Masked node features.

Return type:

torch.Tensor