pyNIBS 0.2024.8__py3-none-any.whl

pynibs/expio/fit_funs.py

@@ -0,0 +1,287 @@
+import numpy as np
+
+
+def dummy_fun(x, a):
+    """
+    Dummy function for congruence factor calculation.
+    """
+    return x
+
+
+def sigmoid(x, x0, r, amp):
+    """
+    Parametrized sigmoid function.
+
+    .. math::
+        y = \\frac{amp}{1+e^{-r(x-x_0)}}
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+    amp : float
+        Maximum value the sigmoid converges to.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at argument x.
+    """
+    y = amp / (1 + np.exp(-r * (x - x0)))
+    return y
+
+
+def sigmoid4(x, x0, r, amp, y0):
+    """
+    Parametrized sigmoid function with 4 parameters.
+
+    .. math::
+        y = y_0 + \\frac{amp - y_0}{1+e^{-r(x-x_0)}}
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) x-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+    amp : float
+        Maximum value the sigmoid converges to.
+    y0 : float
+        Offset value of the sigmoid.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at argument x.
+    """
+    exp = -r * (x - x0)
+
+    # avoid numpy RuntimeWarning: overflow encountered in exp
+    if np.all(exp < 500):
+        y = y0 + (amp - y0) / (1 + np.exp(exp))
+    else:
+        # return a flat line; high value to maximally impair fits
+        y = np.ones(x.shape)
+        y *= np.iinfo(np.int32).max
+
+    return y
+
+
+def sigmoid_log(x, x0, r, amp):
+    """
+    Parametrized log transformed sigmoid function.
+
+    .. math::
+        y = \\log\\left(\\frac{amp}{1+e^{-r(x-x_0)}}\\right)
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) x-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+    amp : float
+        Maximum value the sigmoid converges to.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at argument x.
+    """
+    y = np.log10(1e-12 + (amp - 1e-12) / (1 + np.exp(-r * (x - x0))))
+    return y
+
+
+def sigmoid4_log(x, x0, r, amp, y0):
+    """
+    Parametrized log transformed sigmoid function with 4 parameters.
+
+    .. math::
+        y = \\log\\left(y_0 + \\frac{amp - y_0}{1+e^{-r(x-x_0)}}\\right)
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+    amp : float
+        Maximum value the sigmoid converges to (upper saturation).
+    y0 : float
+        Y-offset value of the sigmoid.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at argument x.
+    """
+    return np.log10(sigmoid4(y0=y0, amp=amp, x=x, x0=x0, r=r))
+
+
+def linear(x, m, n):
+    """
+    Parametrized linear function.
+
+    .. math::
+        y = mx+n
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    m : float
+        Slope parameter.
+    n : float
+        Y-offset.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at argument x.
+    """
+    y = m * x + n
+    return y
+
+
+def linear_log(x, m, n):
+    """
+    Parametrized log linear function.
+
+    .. math::
+        y = mx+n
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    m : float
+        Slope parameter (steepness).
+    n : float
+        Y-offset.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at x.
+    """
+    y = m * x + n
+    y[y <= 0] = 1e-16
+    y = np.log10(y)
+    y[y == -np.inf] = -16
+    y[y == np.inf] = 16
+
+    return y
+
+
+def exp(x, x0, r, y0):
+    """
+    Parametrized exponential function.
+
+    .. math::
+        y = y_0 + e^{r(x-x_0)}
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+    y0: float
+        Offset parameter.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x,) Function value at x.
+    """
+    y = y0 + np.exp(r * (x - x0))
+    return y
+
+
+def exp_log(x, x0, r, y0):
+    """
+    Parametrized exponential function (log).
+
+    .. math::
+        y = y_0 + e^{r(x-x_0)}
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+    y0: float
+        y-offset parameter.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at x.
+    """
+    y = np.log10(y0 + np.exp(r * (x - x0)))
+    return y
+
+
+def exp0(x, x0, r):
+    """
+    Parametrized exponential function w/o offset.
+
+    .. math::
+        y = e^{r(x-x_0)}
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter (steepness).
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at x.
+    """
+    y = np.exp(r * (x - x0))
+    return y
+
+
+def exp0_log(x, x0, r):
+    """
+    Parametrized exponential function w/o offset.
+
+    .. math::
+        y = e^{r(x-x_0)}
+
+    Parameters
+    ----------
+    x : np.ndarray of float
+        (N_x) X-values the function is evaluated in.
+    x0 : float
+        Horizontal shift along the abscissa.
+    r : float
+        Slope parameter.
+
+    Returns
+    -------
+    y : np.ndarray of float
+        (N_x) Function value at x.
+    """
+    y = np.log10(np.exp(r * (x - x0)))
+    return y