Source code for oats.models.predictive.tcn

"""
Temporal Convolution Networks (TCN)
-----------------
"""
from typing import Any
from functools import partial

from darts import models
import numpy as np
import numpy.typing as npt
import optuna

from oats.models._darts_model import DartsModel


class TCNModel(DartsModel):
    """TCN Model (Temporal Convolution Network)

    Using TCN as a predictor. Anomalies scores are deviations from predictions.

    Reference: https://unit8co.github.io/darts/generated_api/darts.models.forecasting.tcn_model.html
    """

    def __init__(
        self,
        window: int = 10,
        n_steps: int = 1,
        use_gpu: bool = 1,
        val_split: float = 0.2,
        **kwargs
    ):
        """
        initialization also accepts any parameters used by: https://unit8co.github.io/darts/generated_api/darts.models.forecasting.tcn_model.html

        Args:
            window (int, optional): rolling window size to feed into the predictor. Defaults to 10.
            n_steps (int, optional): number of steps to predict forward. Defaults to 1.
            use_gpu (bool, optional): whether to use GPU. Defaults to False.
            val_split (float, optional): proportion of data points reserved for validation. Defaults to 0.2.
        """

        model = models.TCNModel

        super().__init__(model, window, n_steps, use_gpu, val_split, **kwargs)

    def _model_objective(self, trial, train_data: npt.NDArray[Any]):
        params = {
            "kernel_size": trial.suggest_int(
                "kernel_size", 2, min(32, self.window - 1)
            ),
            "num_filters": trial.suggest_int("num_filters", 2, 8),
            "weight_norm": trial.suggest_categorical("weight_norm", [True, False]),
            "dilation_base": trial.suggest_int("dilation_base", 1, 4),
            "dropout": trial.suggest_float("dropout", 0.0, 0.3),
        }

        return self._get_hyperopt_res(params, train_data)