scSLAT.model.utils.run_SLAT

scSLAT.model.utils.run_SLAT(features, edges, epochs=6, LGCN_layer=1, mlp_hidden=256, hidden_size=2048, alpha=0.01, anchor_scale=0.8, lr_mlp=0.0001, lr_wd=0.0001, lr_recon=0.01, batch_d_per_iter=5, batch_r_per_iter=10)

Run SLAT model

Parameters:

• features (typing.List) – list of graph node features

• edges (typing.List) – list of graph edges

• epochs (typing.Optional[int]) – epoch number of SLAT (not exceed 10)

• LGCN_layer (typing.Optional[int]) – LGCN layer number, we suggest set 1 for barcode based and 4 for fluorescence based

• mlp_hidden (typing.Optional[int]) – MLP hidden layer size

• hidden_size (typing.Optional[int]) – size of LGCN output

• transform – if use transform

• alpha (typing.Optional[float]) – scale of loss

• anchor_scale (typing.Optional[float]) – ratio of cells selected as pairs

• lr_mlp (typing.Optional[float]) – learning rate of MLP

• lr_wd (typing.Optional[float]) – learning rate of WGAN discriminator

• lr_recon (typing.Optional[float]) – learning rate of reconstruction

• batch_d_per_iter (typing.Optional[int]) – batch number for WGAN train per iter

• batch_r_per_iter (typing.Optional[int]) – batch number for reconstruct train per iter

Return type:

typing.List

Returns:

• embd0 – cell embedding of dataset1

• embd1 – cell embedding of dataset2

• time – run time of SLAT model