Module likelihood.models.utils

Classes

class FeaturesArima

Expand source code

class FeaturesArima:
    """Base class for ARIMA feature computations."""

    def forward(self, y_sum: np.ndarray, theta: list, mode: bool, noise: float) -> np.ndarray:
        """Compute the forward pass of the model.

        Parameters
        ----------
        y_sum : `np.ndarray`
            Summed data points.
        theta : `list`
            Model parameters.
        mode : `bool`
            Forward pass mode.
        noise : `float`
            Noise level.

        Returns
        -------
        `np.ndarray`
            Computed values.
        """
        if mode:
            y_vec = []

            y_t = np.dot(theta, y_sum)

            n = y_sum.shape[0]

            for i in range(n):
                try:
                    n_int = np.where(y_sum != y_sum[i])[0]
                    y_i = (y_t - np.dot(theta[n_int], y_sum[n_int])) / theta[i]
                    y_i += np.random.rand() * noise
                except:
                    y_i = (y_t - np.dot(theta[0:i], y_sum[0:i])) / theta[i]
                y_vec.append(y_i)
        else:
            y_t = np.dot(theta, y_sum) + y_sum[0]
            n_int = np.where(y_sum != y_sum[0])[0]
            y_i = (y_t - np.dot(theta[n_int], y_sum[n_int])) / theta[0]
            y_i += np.random.rand() * noise
            return y_i

        return np.array(y_vec)

    def integrated(self, datapoints: np.ndarray) -> np.ndarray:
        """Compute integrated values.

        Parameters
        ----------
        datapoints : `np.ndarray`
            Input data points.

        Returns
        -------
        `np.ndarray`
            Integrated values.
        """
        datapoints = self.datapoints
        y_sum = list(np.diff(datapoints, self.d))
        y_sum.insert(0, datapoints[0])

        return np.array(y_sum)

    def average(self, datapoints: np.ndarray) -> np.ndarray:
        """Compute average values.

        Parameters
        ----------
        datapoints : `np.ndarray`
            Input data points.

        Returns
        -------
        `np.ndarray`
            Average values.
        """
        y_sum_average = cal_average(datapoints)
        y_sum_eps = datapoints - y_sum_average

        return y_sum_eps

Base class for ARIMA feature computations.

Subclasses

• AbstractArima

Methods

def average(self, datapoints: numpy.ndarray) ‑> numpy.ndarray

Expand source code

def average(self, datapoints: np.ndarray) -> np.ndarray:
    """Compute average values.

    Parameters
    ----------
    datapoints : `np.ndarray`
        Input data points.

    Returns
    -------
    `np.ndarray`
        Average values.
    """
    y_sum_average = cal_average(datapoints)
    y_sum_eps = datapoints - y_sum_average

    return y_sum_eps

Compute average values.

Parameters

datapoints : np.ndarray

Input data points.

Returns

np.ndarray Average values.

def forward(self, y_sum: numpy.ndarray, theta: list, mode: bool, noise: float) ‑> numpy.ndarray

Expand source code

def forward(self, y_sum: np.ndarray, theta: list, mode: bool, noise: float) -> np.ndarray:
    """Compute the forward pass of the model.

    Parameters
    ----------
    y_sum : `np.ndarray`
        Summed data points.
    theta : `list`
        Model parameters.
    mode : `bool`
        Forward pass mode.
    noise : `float`
        Noise level.

    Returns
    -------
    `np.ndarray`
        Computed values.
    """
    if mode:
        y_vec = []

        y_t = np.dot(theta, y_sum)

        n = y_sum.shape[0]

        for i in range(n):
            try:
                n_int = np.where(y_sum != y_sum[i])[0]
                y_i = (y_t - np.dot(theta[n_int], y_sum[n_int])) / theta[i]
                y_i += np.random.rand() * noise
            except:
                y_i = (y_t - np.dot(theta[0:i], y_sum[0:i])) / theta[i]
            y_vec.append(y_i)
    else:
        y_t = np.dot(theta, y_sum) + y_sum[0]
        n_int = np.where(y_sum != y_sum[0])[0]
        y_i = (y_t - np.dot(theta[n_int], y_sum[n_int])) / theta[0]
        y_i += np.random.rand() * noise
        return y_i

    return np.array(y_vec)

Compute the forward pass of the model.

Parameters

y_sum : np.ndarray

Summed data points.

theta : list

Model parameters.

mode : bool

Forward pass mode.

noise : float

Noise level.

Returns

np.ndarray Computed values.

def integrated(self, datapoints: numpy.ndarray) ‑> numpy.ndarray

Expand source code

def integrated(self, datapoints: np.ndarray) -> np.ndarray:
    """Compute integrated values.

    Parameters
    ----------
    datapoints : `np.ndarray`
        Input data points.

    Returns
    -------
    `np.ndarray`
        Integrated values.
    """
    datapoints = self.datapoints
    y_sum = list(np.diff(datapoints, self.d))
    y_sum.insert(0, datapoints[0])

    return np.array(y_sum)

Compute integrated values.

Parameters

datapoints : np.ndarray

Input data points.

Returns

np.ndarray Integrated values.