BiTCN
Bidirectional Temporal Convolutional Network (BiTCN) is a forecasting architecture based on two temporal convolutional networks (TCNs). The first network (‘forward’) encodes future covariates of the time series, whereas the second network (‘backward’) encodes past observations and covariates. This method allows to preserve the temporal information of sequence data, and is computationally more efficient than common RNN methods (LSTM, GRU, …). As compared to Transformer-based methods, BiTCN has a lower space complexity, i.e. it requires orders of magnitude less parameters.
This model may be a good choice if you seek a small model (small amount of trainable parameters) with few hyperparameters to tune (only 2).
References
-Olivier Sprangers, Sebastian Schelter, Maarten de
Rijke (2023). Parameter-Efficient Deep Probabilistic Forecasting.
International Journal of Forecasting 39, no. 1 (1 January 2023): 332–45.
URL:
https://doi.org/10.1016/j.ijforecast.2021.11.011.
-Shaojie Bai, Zico Kolter, Vladlen Koltun. (2018). An Empirical
Evaluation of Generic Convolutional and Recurrent Networks for Sequence
Modeling. Computing Research Repository, abs/1803.01271. URL:
https://arxiv.org/abs/1803.01271.
-van den Oord, A., Dieleman, S., Zen, H., Simonyan, K., Vinyals, O.,
Graves, A., Kalchbrenner, N., Senior, A. W., & Kavukcuoglu, K. (2016).
Wavenet: A generative model for raw audio. Computing Research
Repository, abs/1609.03499. URL: http://arxiv.org/abs/1609.03499.
arXiv:1609.03499.
1. Auxiliary Functions
source
TCNCell
Temporal Convolutional Network Cell, consisting of CustomConv1D modules.
source
CustomConv1d
Forward- and backward looking Conv1D
2. BiTCN
source
BiTCN
*BiTCN
Bidirectional Temporal Convolutional Network (BiTCN) is a forecasting architecture based on two temporal convolutional networks (TCNs). The first network (‘forward’) encodes future covariates of the time series, whereas the second network (‘backward’) encodes past observations and covariates. This is a univariate model.
Parameters:
h: int, forecast horizon.
input_size: int,
considered autorregresive inputs (lags), y=[1,2,3,4] input_size=2 ->
lags=[1,2].
hidden_size: int=16, units for the TCN’s hidden
state size.
dropout: float=0.1, dropout rate used for the dropout
layers throughout the architecture.
futr_exog_list: str list,
future exogenous columns.
hist_exog_list: str list, historic
exogenous columns.
stat_exog_list: str list, static exogenous
columns.
exclude_insample_y: bool=False, the model skips the
autoregressive features y[t-input_size:t] if True.
loss: PyTorch
module, instantiated train loss class from losses
collection.
valid_loss: PyTorch module=loss, instantiated valid loss class from
losses
collection.
max_steps: int=1000, maximum number of training steps.
learning_rate: float=1e-3, Learning rate between (0, 1).
num_lr_decays: int=-1, Number of learning rate decays, evenly
distributed across max_steps.
early_stop_patience_steps: int=-1,
Number of validation iterations before early stopping.
val_check_steps: int=100, Number of training steps between every
validation loss check.
batch_size: int=32, number of different
series in each batch.
valid_batch_size: int=None, number of
different series in each validation and test batch, if None uses
batch_size.
windows_batch_size: int=1024, number of windows to
sample in each training batch, default uses all.
inference_windows_batch_size: int=-1, number of windows to sample in
each inference batch, -1 uses all.
start_padding_enabled:
bool=False, if True, the model will pad the time series with zeros at
the beginning, by input size.
step_size: int=1, step size between
each window of temporal data.
scaler_type: str=‘identity’, type of
scaler for temporal inputs normalization see temporal
scalers.
random_seed: int=1, random_seed for pytorch initializer and numpy
generators.
num_workers_loader: int=os.cpu_count(), workers to be
used by TimeSeriesDataLoader.
drop_last_loader: bool=False, if
True TimeSeriesDataLoader drops last non-full batch.
alias: str,
optional, Custom name of the model.
optimizer: Subclass of
‘torch.optim.Optimizer’, optional, user specified optimizer instead of
the default choice (Adam).
optimizer_kwargs: dict, optional, list
of parameters used by the user specified optimizer.
lr_scheduler: Subclass of ‘torch.optim.lr_scheduler.LRScheduler’,
optional, user specified lr_scheduler instead of the default choice
(StepLR).
lr_scheduler_kwargs: dict, optional, list of parameters
used by the user specified lr_scheduler.
**trainer_kwargs: int,
keyword trainer arguments inherited from PyTorch Lighning’s
trainer.
References
BiTCN.fit
*Fit.
The fit method, optimizes the neural network’s weights using the
initialization parameters (learning_rate, windows_batch_size, …) and
the loss function as defined during the initialization. Within fit
we use a PyTorch Lightning Trainer that inherits the initialization’s
self.trainer_kwargs, to customize its inputs, see PL’s trainer
arguments.
The method is designed to be compatible with SKLearn-like classes and in particular to be compatible with the StatsForecast library.
By default the model is not saving training checkpoints to protect
disk memory, to get them change enable_checkpointing=True in
__init__.
Parameters:
dataset: NeuralForecast’s
TimeSeriesDataset,
see
documentation.
val_size: int, validation size for temporal cross-validation.
random_seed: int=None, random_seed for pytorch initializer and numpy
generators, overwrites model.__init__’s.
test_size: int, test
size for temporal cross-validation.
*
BiTCN.predict
*Predict.
Neural network prediction with PL’s Trainer execution of
predict_step.
Parameters:
dataset: NeuralForecast’s
TimeSeriesDataset,
see
documentation.
test_size: int=None, test size for temporal cross-validation.
step_size: int=1, Step size between each window.
random_seed:
int=None, random_seed for pytorch initializer and numpy generators,
overwrites model.__init__’s.
**data_module_kwargs: PL’s
TimeSeriesDataModule args, see
documentation.*