scCMA

scCMA: a constrasive masked autoencoder for single-cell RNA-seq clustering

Datasets

Place the h5 dataset in the data/ directory, h5 file contains gene expression X and true label Y.

Usage

Set the dataset read path and result storage path

# Datasets directory and output directory
args["paths"] = {"data": "./data/", "results": "./result/"} 

Parameter settings.

hidden_size=128,
dropout=0.1,
masked_data_weight=0.75,
mask_loss_weight=0.7,
contrastive_weight=0.07

scCMA can be run with the following commands.

python run_scCMA.py