Unsupervised graph-based loss

Sec. 3.2 of Hamilton et al. 2017: nearby nodes agree, sampled negatives disagree.

Usage

morie_gsageemd_loss(z_u, z_v, z_negatives)

Arguments

z_u

Anchor embedding.

z_v

Positive (neighbour) embedding.

z_negatives

List of negative embeddings.

Value

Scalar loss.

References

Hamilton, W. L. et al. (2017).

Examples

morie_gsageemd_loss(z_u = c(1, 0), z_v = c(0.9, 0.1), z_negatives = list(c(-1, 0), c(0, -1)))
#> [1] 1.347563