scpdac.tl.MLP

Contents

scpdac.tl.MLP#

class scpdac.tl.MLP(n_genes, n_classes, hidden=(512, 256), dropout=0.3)#

Bases: Module

A simple feed-forward classifier with BatchNorm + dropout hidden blocks.

Parameters:

  • n_genes (int) – Number of input features (genes).

  • n_classes (int) – Number of output classes.

  • hidden (tuple[int, ...] (default: (512, 256))) – Sizes of the hidden layers.

  • dropout (float (default: 0.3)) – Dropout probability applied after each hidden block.

forward(x)#

Return raw class logits for a (n_cells, n_genes) input tensor.

Return type:

Tensor

Contents