Repository files navigation

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

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

Set the dataset read path and result storage path

# Datasets directory and output directoryargs["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