sc_reconstruction.decoders.ReconMLPDecoder#

class sc_reconstruction.decoders.ReconMLPDecoder(n_input, n_output, n_cat_list=None, n_layers=1, n_hidden=128, distribution='normal', learning_rate=0.001, reduce_lr_on_plateau=False, lr_factor=0.6, lr_patience=5, lr_threshold=0.001, lr_min=0.0, library_size_mode='none', decoder_output_activation=None, **kwargs)[source]#

Bases: BaseReconstructionDecoder

Parameters:

  • n_input (int)

  • n_output (int)

  • n_cat_list (Iterable[int])

  • n_layers (int)

  • n_hidden (int)

  • distribution (str)

  • learning_rate (float)

  • reduce_lr_on_plateau (bool)

  • lr_factor (float)

  • lr_patience (int)

  • lr_threshold (float)

  • lr_min (float)

  • library_size_mode (str)

  • decoder_output_activation (str)

__init__(n_input, n_output, n_cat_list=None, n_layers=1, n_hidden=128, distribution='normal', learning_rate=0.001, reduce_lr_on_plateau=False, lr_factor=0.6, lr_patience=5, lr_threshold=0.001, lr_min=0.0, library_size_mode='none', decoder_output_activation=None, **kwargs)[source]#
Parameters:

Methods

__init__(n_input, n_output[, n_cat_list, ...])

decode(z, *cat_list)

Decode latent representations to data space

load(path[, map_location])

Load the model weights

save(path)

Save the trained decoder.

train([datamodule])