scglue.models.sc.DataEncoder.forward

DataEncoder.forward(x, xalt, lazy_normalizer=True)

Encode data to sample latent distribution

Parameters

• x (Tensor) – Input data

• xalt (Tensor) – Alternative input data

• lazy_normalizer (bool) – Whether to skip computing x normalizer (just return None) if xalt is non-empty

Return type

Tuple[Normal, Optional[Tensor]]

Returns

• u – Sample latent distribution

• normalizer – Data normalizer

Note

Normalization is always computed on x. If xalt is empty, the normalized x will be used as input to the encoder neural network, otherwise xalt is used instead.