Fast single cell clustering parameters

Usage

params_sc_fast_cluster(
  kmeans_iters = 100L,
  batch_size = 4096L,
  drift_threshold = 1e-04,
  lr_alpha = 1,
  full_snn = FALSE,
  pruning = NULL,
  snn_similarity = c("jaccard", "rank"),
  louvain_iters = 10L,
  knn = list(k = 5L)
)

Arguments

kmeans_iters: Integer. Number of iterations for k-means clustering. Defaults to 100L.
batch_size: Integer. Batch size for mini batch k-means clustering. Defaults to 4096L.
drift_threshold: Numeric. The drift for the mini batch k-means clustering. If the centroid drift is below this, the mini batch k-means terminates. Defaults to 1e-04.
lr_alpha: Numeric. Learning rate alpha parameter for mini batch k-means. Defaults to 1.0.
full_snn: Boolean. Shall the full shared nearest neighbour graph be generated that generates edges between all cells instead of between only neighbours. Defaults to FALSE.
pruning: Numeric or NULL. Weights below this threshold will be set to 0 in the generation of the sNN graph. If not provided, defaults to 1 / ceil(k * 0.8). Defaults to NULL.
snn_similarity: String. The Jaccard similarity calculates the Jaccard between the neighbours, whereas the rank method calculates edge weights based on the ranking of shared neighbours. For the rank method, the weight is determined by finding the shared neighbour with the lowest combined rank across both cells, where lower-ranked (closer) shared neighbours result in higher edge weights Both methods produce weights normalised to the range [0, 1]. One of c("jaccard", "rank"). Defaults to "jaccard".
louvain_iters: Integer. Number of iterations for Louvain clustering. Defaults to 10L.
knn: List. Optional overrides for kNN parameters. See params_knn_defaults() for available parameters: k, knn_method, ann_dist, search_budget, n_trees, delta, diversify_prob, ef_budget, extract_knn, m, ef_construction, ef_search, n_list and n_probe. Sets the default k = 5L. See params_knn_defaults() for the available elements. Defaults to list(k = 5L).

Value

A named list with the following elements:

kmeans_iters - Integer. Number of iterations for k-means clustering. Defaults to 100L.
batch_size - Integer. Batch size for mini batch k-means clustering. Defaults to 4096L.
drift_threshold - Numeric. The drift for the mini batch k-means clustering. If the centroid drift is below this, the mini batch k-means terminates. Defaults to 1e-04.
lr_alpha - Numeric. Learning rate alpha parameter for mini batch k-means. Defaults to 1.0.
louvain_iters - Integer. Number of iterations for Louvain clustering. Defaults to 10L.
full_snn - Boolean. Shall the full shared nearest neighbour graph be generated that generates edges between all cells instead of between only neighbours. Defaults to FALSE.
pruning - Numeric or NULL. Weights below this threshold will be set to 0 in the generation of the sNN graph. If not provided, defaults to 1 / ceil(k * 0.8). Defaults to NULL.
snn_similarity - String. The Jaccard similarity calculates the Jaccard between the neighbours, whereas the rank method calculates edge weights based on the ranking of shared neighbours. For the rank method, the weight is determined by finding the shared neighbour with the lowest combined rank across both cells, where lower-ranked (closer) shared neighbours result in higher edge weights Both methods produce weights normalised to the range [0, 1]. One of c("jaccard", "rank"). Defaults to "jaccard".
The elements of params_knn_defaults(), overridden by knn, spliced in at this position.