Skip to contents

Triplet Loss-Based Metric Learning for Single-Cell Data

Source: R/scLearn_model_learning.R
runLMNN.Rd

Implements a triplet loss-based linear transformation learning to map gene expression into an embedding space where cells of the same type are close and those of different types are farther apart. This is a deep metric learning alternative to LMNN.

Usage

runLMNN(
  high_varGenes,
  expression_profile,
  sample_information,
  k = 5,
  margin = 1,
  learn_rate = 1e-04,
  max_iter = 100,
  seed = 1,
  verbose = TRUE
)

Arguments

high_varGenes

Character vector of highly variable gene names.

expression_profile

A numeric matrix of gene expression values (genes x cells).

sample_information

A named vector of cell type labels (names match column names of expression_profile).

k

Number of positive neighbors per anchor to form triplets (default: 5).

margin

Margin between positive and negative pairs (default: 1.0).

learn_rate

Learning rate for updates (default: 1e-4).

max_iter

Number of optimization iterations (default: 100).

seed

Random seed (default: 1).

verbose

Logical; whether to print progress messages (default: TRUE).

Value

A list containing:

  • expression_profile_trans: Transformed expression matrix (genes x cells).

  • expression_profile_origin: Original filtered expression matrix (genes x cells).

  • trans_matrix: Learned linear transformation matrix (L).

  • sample_information: Sorted cell type labels.