Feature selection based on deep neural network for scRNAseq data
In order to run scFSNN, an operative version of Python and TensorFlow is needed. The code has been tested on Python 3.8.12 and PyTorch 1.10.2. Other Python dependencies include numpy and scikit-learn.
scFSNN is used to select features based on deep neural network for scRNAseq data.
scFSNN(train_X, train_Y, val_X, val_Y, test_X, test_Y, num_classes, lr, epochs, batch_size, q0, device, dropout=0.5, eta=1, elimination_rate=1, cut_off=0.1)
- train_X training inputs
- train_Y training output
- val_X validation inputs
- val_Y validation output
- test_X test inputs
- test_Y test output
- num_classes number of classes of output
- lr learning rate
- epochs number of epochs to train
- q0 number of dummy variables
- device "gpu" or "cpu"
- dropout dropout rate to control the overfitting