rllm.transforms.graph_transforms.KNNGraph

class rllm.transforms.graph_transforms.KNNGraph(num_neighbors: int | None = 6, mode: str | None = 'connectivity', metric: str | None = 'minkowski', p: int | None = 2, metric_params: dict | None = None, include_self: bool | None = False, n_jobs: int = 1)[source]

Bases: EdgeTransform

Builds a k-NN adjacency matrix from node features.

Parameters:

  • num_neighbors (int, optional) – The number of neighbors. (default: 6)

  • mode (str[connectivity, distance], optional) – Type of returned matrix: connectivity will return the connectivity matrix with ones and zeros, while distance will return the distances between neighbors according to the given metric. (default: connectivity)

  • metric (str[minkowski, cosine, l1, l2, …], optional) – Metric to use for distance computation. Default is minkowski, which results in the standard Euclidean distance when p = 2. (default: minkowski)

  • p (float) – Power parameter for the Minkowski metric (default: 2).

  • metric_params (dict, optional) – Additional keyword arguments for the metric function. (default: None)

  • include_self (bool, optional) – If set to True, the graph will contain self-loops. (default: False)

  • n_jobs (int) – Number of workers to use for computation. (default: 1)