paddlets.models.anomaly.dl._ed.ed

class MLP(input_dim: int, feature_dim: int, hidden_config: List[int], activation: Callable[[...], Tensor], last_layer_activation: Callable[[...], Tensor], dropout_rate: float = 0.5, use_bn: bool = True, use_drop: bool = True)[source]

Bases: Layer

MLP Network structure used in the encoder and decoder

Parameters

  • input_dim (int) – The size of the loopback window, i.e. the number of time steps feed to the model.

  • feature_dim (int) – The numer of feature.

  • hidden_config (List(int)) – The ith element represents the number of neurons in the ith hidden layer.

  • activation (Callable[..., paddle.Tensor]) – The activation function for the hidden layers.

  • last_layer_activation (Callable[..., paddle.Tensor]) – The activation function for the last layer.

  • dropout_rate (float) – Dropout regularization parameter.

  • use_bn (bool) – Whether to use batch normalization.

  • use_drop (bool) – Whether to use dropout.

_nn

Dynamic graph LayerList.

Type

paddle.nn.Sequential

forward(x)[source]

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

  • *inputs (tuple) – unpacked tuple arguments

  • **kwargs (dict) – unpacked dict arguments

class CNN(input_dim: int, hidden_config: List[int], activation: Callable[[...], Tensor], last_layer_activation: Callable[[...], Tensor], kernel_size: int, dropout_rate: float = 0.5, use_bn: bool = True, is_encoder: bool = True, use_drop: bool = True, data_format: str = 'NCL')[source]

Bases: Layer

CNN Network structure used in the encoder and decoder

Parameters

  • input_dim (int) – The size of the loopback window, i.e. the number of time steps feed to the model.

  • hidden_config (List(int)) – The ith element represents the number of neurons in the ith hidden layer.

  • activation (Callable[..., paddle.Tensor]) – The activation function for the hidden layers.

  • last_layer_activation (Callable[..., paddle.Tensor]) – The activation function for the last layer.

  • kernel_size (int) – Kernel size for Conv1D.

  • use_drop (bool) – Whether to use dropout.

  • dropout_rate (float) – Dropout regularization parameter.

  • use_bn (bool) – Whether to use batch normalization.

  • is_encoder (bool) – Encoder or Decoder.

  • data_format (str) – Specify the input data format.N is the batch size, C is the number of channels and L is the characteristic length.

_nn

Dynamic graph LayerList.

Type

paddle.nn.Sequential

forward(x)[source]

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

  • *inputs (tuple) – unpacked tuple arguments

  • **kwargs (dict) – unpacked dict arguments

class LSTM(input_dim: int, hidden_config: List[int], activation: Callable[[...], Tensor], last_layer_activation: Callable[[...], Tensor], dropout_rate: float = 0, use_drop: bool = True, num_layers: int = 1, direction: str = 'forward')[source]

Bases: Layer

LSTM Network structure used in the encoder and decoder

Parameters

  • input_dim (int) – The size of the loopback window, i.e. the number of time steps feed to the model.

  • hidden_config (List(int)) – The ith element represents the number of neurons in the ith hidden layer.

  • activation (Callable[..., paddle.Tensor]) – The activation function for the hidden layers.

  • last_layer_activation (Callable[..., paddle.Tensor]) – The activation function for the last layer.

  • dropout_rate (float) – Dropout regularization parameter.

  • use_drop (bool) – Whether to use dropout.

  • num_layers (int) – layers of LSTM.

  • direction (str) – the direction of LSTM.

_nn

Dynamic graph LayerList.

Type

paddle.nn.Sequential

forward(x)[source]

Defines the computation performed at every call. Should be overridden by all subclasses.

Parameters

  • *inputs (tuple) – unpacked tuple arguments

  • **kwargs (dict) – unpacked dict arguments