Back to models
Classifier

GRUFCNNClassifier

GRU-FCN for time series classification.

Quickstart

python
from sktime.classification.deep_learning.gru import GRUFCNNClassifier

estimator = GRUFCNNClassifier(hidden_dim: int, gru_layers: int, batch_first: bool=False, bias: bool=True, init_weights: bool=True, dropout: float=0.0, gru_dropout: float=0.0, bidirectional: bool=False, conv_layers: list=[128, 256, 128], kernel_sizes: list=[7, 5, 3], num_epochs: int=10, batch_size: int=8, optimizer: str='Adam', criterion: str=None, criterion_kwargs: dict=None, optimizer_kwargs: dict=None, lr: float=0.01, verbose: bool=False, random_state: int=None)

Parameters(20)

hidden_dimint
Number of features in the hidden state.
gru_layersint
Number of recurrent layers.
batch_firstbool
If True, then the input and output tensors are provided as (batch, seq, feature), default is False.
biasbool
If False, then the layer does not use bias weights, default is True.
init_weightsbool
If True, then the weights are initialized, default is True.
dropoutfloat
Dropout rate to apply inside gru cell. default is 0.0
gru_dropoutfloat
Dropout rate to apply to the gru output layer. default is 0.0
bidirectionalbool
If True, then the GRU is bidirectional, default is False.
conv_layerslist
List of integers specifying the number of filters in each convolutional layer. default is [128, 256, 128].
kernel_sizeslist
List of integers specifying the kernel size in each convolutional layer. default is [7, 5, 3].
num_epochsint, optional (default=10)
The number of epochs to train the model.
optimizerstr, optional (default=None)
The optimizer to use. If None, Adam will be used.
activationstr, optional (default=”relu”)
The activation function to use. Options: [“relu”, “softmax”].
batch_sizeint, optional (default=8)
The size of each mini-batch during training.
criterioncallable, optional (default=None)
The loss function to use. If None, CrossEntropyLoss will be used.
criterion_kwargsdict, optional (default=None)
Additional keyword arguments to pass to the loss function.
optimizer_kwargsdict, optional (default=None)
Additional keyword arguments to pass to the optimizer.
lrfloat, optional (default=0.001)
The learning rate to use for the optimizer.
verbosebool, optional (default=False)
Whether to print progress information during training.
random_stateint, optional (default=None)
Seed to ensure reproducibility.

References

  1. [1 ] Elsayed, et al. “Deep Gated Recurrent and Convolutional Network Hybrid Model for Univariate Time Series Classification.” arXiv preprint arXiv:1812.07683 (2018). [2 ] https://github.com/NellyElsayed/GRU-FCN-model-for-univariate-time-series-classification [3 ] https://github.com/titu1994/LSTM-FCN