AEAttentionBiGRUNetwork

class AEAttentionBiGRUNetwork(latent_space_dim=None, temporal_latent_space=False, n_layers_encoder=1, n_layers_decoder=1, activation_encoder='relu', activation_decoder='relu')

Bases: BaseDeepLearningNetwork

A class to implement an Auto-Encoder based on Attention Bidirectional GRUs.

Parameters:

latent_space_dimint, default=128

Dimension of the latent space.

temporal_latent_spacebool, default=False

Flag to choose whether the latent space is an MTS or Euclidean space.

n_layers_encoderint, default=None

Number of Attention BiGRU layers in the encoder. If None, one layer will be used.

n_layers_decoderint, default=None

Number of Attention BiGRU layers in the decoder. If None, one layer will be used.

activation_encoderUnion[list, str], default=”relu”

Activation function(s) to use in each layer of the encoder. Can be a single string or a list.

activation_decoderUnion[list, str], default=”relu”

Activation function(s) to use in each layer of the decoder. Can be a single string or a list.

References

[1]

Ienco, D., & Interdonato, R. (2020). Deep multivariate time series

embedding clustering via attentive-gated autoencoder. In Advances in Knowledge Discovery and Data Mining: 24th Pacific-Asia Conference, PAKDD 2020, Singapore, May 11-14, 2020, Proceedings, Part I 24 (pp. 318-329). Springer International Publishing.

Methods

build_network(input_shape, **kwargs)

Construct a network and return its input and output layers.

build_network(input_shape, **kwargs)

Construct a network and return its input and output layers.

Parameters:

input_shapetuple of shape = (n_timepoints (m), n_channels (d))

The shape of the data fed into the input layer.

kwargsdict

Contains additional keyword arguments to be passed to the build_network method like num_input_samples.

Returns:

encodera keras Model.

decodera keras Model.