SemiSupervisedTrainer

class scvi.inference.SemiSupervisedTrainer(model, gene_dataset, n_labelled_samples_per_class=50, n_epochs_classifier=1, lr_classification=0.005, classification_ratio=50, seed=0, **kwargs)

Bases: scvi.inference.inference.UnsupervisedTrainer

Class for the semi-supervised training of an autoencoder.

This parent class can be inherited to specify the different training schemes for semi-supervised learning

Parameters

n_labelled_samples_per_class – number of labelled samples per class

Attributes Summary

posteriors_loop

Methods Summary

create_posterior([model, gene_dataset, …])
loss(tensors_all, tensors_labelled)
on_epoch_end()

Attributes Documentation

posteriors_loop

Methods Documentation

create_posterior(model=None, gene_dataset=None, shuffle=False, indices=None, type_class=<class 'scvi.inference.annotation.AnnotationPosterior'>)

loss(tensors_all, tensors_labelled)

on_epoch_end()