Missed TensorFlow Dev Summit? Check out the video playlist. Watch recordings

tf_agents.bandits.agents.utils.build_laplacian_nearest_neighbor_graph

View source on GitHub

Build the Laplacian matrix of a nearest neighbor graph.

tf_agents.bandits.agents.utils.build_laplacian_nearest_neighbor_graph(
    input_vecs, k=1
)

Given input embedding vectors, this utility returns the Laplacian matrix of the induced k-nearest-neighbor graph.

Args:

  • input_vecs: a Tensor. Input embedding vectors (one per row). Shaped [num_vectors, ...].
  • k: an integer. Number of nearest neighbors to use.

Returns:

The graph Laplacian matrix. A dense float Tensor of shape [num_vectors, num_vectors], where num_vectors is the number of input embedding vectors (Tensor).