fl4health.model_bases.autoencoders_base module

class AbstractAe(encoder, decoder)

Bases: Module, ABC

__init__(encoder, decoder)

The base class for all autoencoder based models. To define this model, we need to define the structure of the encoder and the decoder modules. This type of model should have the capability to encode data using the encoder module and decode the output of the encoder using the decoder module.

abstract forward(input)

Define the computation performed at every call.

Should be overridden by all subclasses. :rtype: Tensor

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class BasicAe(encoder, decoder)

Bases: AbstractAe

decode(latent_vector)

Return type:

Tensor

encode(input)

Return type:

Tensor

forward(input)

Define the computation performed at every call.

Should be overridden by all subclasses. :rtype: Tensor

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class ConditionalVae(encoder, decoder, unpack_input_condition=None)

Bases: AbstractAe

__init__(encoder, decoder, unpack_input_condition=None)

Conditional Variational Auto-Encoder model.

Parameters:

• encoder (nn.Module) – The encoder used to map input to latent space.

• decoder (nn.Module) – The decoder used to reconstruct the input using a vector in latent space.

• unpack_input_condition (Callable | None, optional) – For unpacking the input and condition tensors.

decode(latent_vector, condition=None)

Return type:

Tensor

encode(input, condition=None)

Return type:

tuple[Tensor, Tensor]

forward(input)

Define the computation performed at every call.

Should be overridden by all subclasses. :rtype: Tensor

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

sampling(mu, logvar)

Return type:

Tensor

class VariationalAe(encoder, decoder)

Bases: AbstractAe

__init__(encoder, decoder)

Variational Auto-Encoder model base class.

Parameters:

• encoder (nn.Module) – Encoder module defined by the user.

• decoder (nn.Module) – Decoder module defined by the user.

decode(latent_vector)

Return type:

Tensor

encode(input)

Return type:

tuple[Tensor, Tensor]

forward(input)

Define the computation performed at every call.

Should be overridden by all subclasses. :rtype: Tensor

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

sampling(mu, logvar)

Return type:

Tensor