Training an ensemble SVM model for embedding coordinates regression.

Training an ensemble SVM model for embedding coordinates regression.

FitEnsembleSVM(
  feature.mat,
  emb.mat,
  cell.types = NULL,
  do.norm = "L2",
  batch.size = 5000,
  n.models = 100,
  balance.cell.type = FALSE,
  cores = -1
)

Arguments

feature.mat

A signature score matrix, rows are cells, columns are features.

emb.mat

The embedding matrix, rows are cells, columns are dimensions.

cell.types

A named vector recording cell types.

do.norm

Whether normalize the feature matrix. L1, L2, NULL. Default: 'L2'

batch.size

The number of cells for each model. Default: 5000

n.models

The number of SVM model. Default: 100

balance.cell.type

A boolen determines whether performing balance sampling. Default: FALSE

cores

The number of CPU for training. Default: -1, use all available threads.

Value

a list of trained learners.