runFeatureAE: Main function of running scGNN feature auto-encoder

In ff98li/scRGNet: Perform scRNA-seq Analysis using Single-cell Graphical Neural Network (scGNN) Framework

Description

The main function for performing dimensional reduction on the preprocessed scRNA-seq data

Usage

runFeatureAE(
  scDataset,
  LTMG_mat = NULL,
  hyperParams = list(batch_size = 1L, regu_epochs = 5L, L1 = 0.5, L2 = 0.5, regu_alpha
    = 0.9, reduction = "sum"),
  hardwareSetup = list(CUDA = F, coresUage = 1L)
)

Arguments

scDataset	A scDataset object containing scRNA-seq data
LTMG_mat	LTMG sparse matrix. If provided, then LTMG regularisation will be applied. Optional.
hyperParams	A list of hyperparameter to tune the model. Optional.
hardwareSetup	A list of parameters to setup the hardware on which the model runs. Optional.

Value

A matrix representing the encoded space of the scRNA-seq matrix.

References

\insertRef

scGNNscRGNet \insertRefLTMGscRGNet \insertReftorchscRGNet \insertRefprogressscRGNet \insertRefmatrixscRGNet

Examples

# Example 1:
# Tested examples. Not run for fast package compiling
## Not run: 
# Accessing the demo gene_counts_small dataset available with the package
inputCountsPath <- system.file("extdata", "GSE138852_small.csv", package = "scRGNet")
# Preprocess the raw counts
counts <- preprocessCSV(path = inputCountsPath)
ltmg <- runLTMG(counts)
hyperParams <- setHyperParams(regu_epochs = 5L)
hardwareSetup <- setHardware(coresUsage = 1L)
z <- runFeatureAE(scDataset = counts, LTMG_mat = ltmg, hyperParams, hardwareSetup)

## End(Not run)

ff98li/scRGNet documentation built on Jan. 14, 2022, 4:58 a.m.