Wrapper function to generate k-means parameters

Wrapper function to generate k-means parameters

Usage

params_kmeans(
  metric = c("euclidean", "cosine"),
  max_iters = 1000L,
  batch_size = 4096L,
  drift_threshold = 1e-04,
  lr_alpha = 1
)

Arguments

metric

Character. Distance metric to use. One of "euclidean" or "cosine". Defaults to "euclidean".

max_iters

Integer. Maximum number of iterations. Defaults to 1000L.

batch_size

Integer. Mini-batch size. Only used when method = "minibatch". Defaults to 4096L.

drift_threshold

Float. Below which centroid drift the mini-batch k-means is considered converged. Defaults to 1e-4. Only used when method = "minibatch"

lr_alpha

Float. Learning rate decay for the mini-batch k-means. Original paper uses 1.0.

Value

A named list with the k-means parameters.