Skip to contents

Helper function to generate kNN defaults

params_knn_defaults.Rd

This function generates various sensible default parameters for all of the different approximate nearest neighbours that are available within this package.

Usage

params_knn_defaults()

Value

A named list with the following elements:

  • k - Integer. Number of neighbours. Defaults to 15L.

  • knn_method - String. Which method to use for the approximate nearest neighbour search. One of c("kmknn", "hnsw", "annoy", "nndescent", "ivf", "exhaustive"). Defaults to "kmknn".

  • ann_dist - String. Which distance metric to use for the approximate nearest neighbour search. One of c("euclidean", "cosine"). Defaults to "euclidean".

  • n_trees - Integer. Annoy param: number of trees to generate for Annoy. Defaults to 50L.

  • search_budget - Integer or NULL. Annoy param: optional search budget per tree for Annoy. If not provided, it will default to n_tree * k * 20L. Defaults to NULL.

  • delta - Numeric. NNDescent param: early termination criterium for NNDescent. Defaults to 0.001.

  • diversify_prob - Numeric. NNDescent param: diversification probability for the NNDescent index. This will diversify the index at the end and identify potentially better edges. Defaults to 0.0.

  • ef_budget - Integer or NULL. NNDescent param: optional query budget parameter. Can accelerate querying, but at the cost of Recall. Defaults to NULL.

  • extract_knn - Boolean. NNDescent param: hand back the graph the descent already built instead of beam searching it. Skips the query pass entirely, so it is much faster, at the cost of some recall. Rows the descent never filled come back padded with duplicate edges. Ignored by every other method. Defaults to FALSE.

  • m - Integer. HNSW param: number of connections between layers for HNSW. Defaults to 16L.

  • ef_construction - Integer. HNSW param: size of dynamic candidate list during construction. Defaults to 200L.

  • ef_search - Integer. HNSW param: size of candidate list (higher = better recall, slower). Defaults to 100L.

  • n_list - Integer or NULL. IVF param: number of clusters/centroids to generate. NULL generates sqrt(n) lists. Defaults to NULL.

  • n_probe - Integer or NULL. IVF param: number of clusters/centroids to query. NULL queries sqrt(n_lists) clusters. Defaults to NULL.