Skip to contents

Wrapper function for the SEACells parameters

params_sc_seacells.Rd

Wrapper function for the SEACells parameters

Usage

params_sc_seacells(
  n_sea_cells,
  max_fw_iters = 50L,
  convergence_epsilon = 0.001,
  max_iter = 100L,
  min_iter = 10L,
  greedy_threshold = 20000L,
  graph_building = "union",
  pruning = TRUE,
  pruning_threshold = 1e-07,
  n_landmarks = NULL,
  knn = list()
)

Arguments

n_sea_cells

Integer. Number of SEA cells to detect. Required.

max_fw_iters

Integer. Maximum iterations for the Franke-Wolfe algorithm. Defaults to 50L.

convergence_epsilon

Numeric. Convergence threshold. Algorithm stops when RSS change < epsilon * RSS(0). Defaults to 0.001.

max_iter

Integer. Maximum iterations to run SEACells for. Defaults to 100L.

min_iter

Integer. Minimum iterations to run SEACells for. Defaults to 10L.

greedy_threshold

Integer. Maximum number of cells before defaulting to rapid random selection of archetypes. Defaults to 20000L.

graph_building

String. Graph building method. Defaults to "union".

pruning

Boolean. Shall tiny values be pruned during Franke-Wolfe updates. Defaults to TRUE.

pruning_threshold

Numeric. If pruning = TRUE values below which threshold shall be pruned. Defaults to 1e-07.

n_landmarks

Integer or NULL. If provided, it will use the Nystroem extension during the archetype finding. Useful for larger data sets. Defaults to NULL.

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. See params_knn_defaults() for the available elements. Defaults to list().

Value

A named list with the following elements:

  • n_sea_cells - Integer. Number of SEA cells to detect. Required.

  • max_fw_iters - Integer. Maximum iterations for the Franke-Wolfe algorithm. Defaults to 50L.

  • convergence_epsilon - Numeric. Convergence threshold. Algorithm stops when RSS change < epsilon * RSS(0). Defaults to 0.001.

  • max_iter - Integer. Maximum iterations to run SEACells for. Defaults to 100L.

  • min_iter - Integer. Minimum iterations to run SEACells for. Defaults to 10L.

  • greedy_threshold - Integer. Maximum number of cells before defaulting to rapid random selection of archetypes. Defaults to 20000L.

  • graph_building - String. Graph building method. Defaults to "union".

  • pruning - Boolean. Shall tiny values be pruned during Franke-Wolfe updates. Defaults to TRUE.

  • pruning_threshold - Numeric. If pruning = TRUE values below which threshold shall be pruned. Defaults to 1e-07.

  • n_landmarks - Integer or NULL. If provided, it will use the Nystroem extension during the archetype finding. Useful for larger data sets. Defaults to NULL.

  • The elements of params_knn_defaults(), overridden by knn, spliced in at this position.