Python/neural_network/activation_functions/soboleva_modified_hyperbolic_tangent.py

49 lines
1.6 KiB
Python
Raw Permalink Normal View History

"""
This script implements the Soboleva Modified Hyperbolic Tangent function.
The function applies the Soboleva Modified Hyperbolic Tangent function
to each element of the vector.
More details about the activation function can be found on:
https://en.wikipedia.org/wiki/Soboleva_modified_hyperbolic_tangent
"""
import numpy as np
def soboleva_modified_hyperbolic_tangent(
vector: np.ndarray, a_value: float, b_value: float, c_value: float, d_value: float
) -> np.ndarray:
"""
Implements the Soboleva Modified Hyperbolic Tangent function
Parameters:
vector (ndarray): A vector that consists of numeric values
a_value (float): parameter a of the equation
b_value (float): parameter b of the equation
c_value (float): parameter c of the equation
d_value (float): parameter d of the equation
Returns:
vector (ndarray): Input array after applying SMHT function
>>> vector = np.array([5.4, -2.4, 6.3, -5.23, 3.27, 0.56])
>>> soboleva_modified_hyperbolic_tangent(vector, 0.2, 0.4, 0.6, 0.8)
array([ 0.11075085, -0.28236685, 0.07861169, -0.1180085 , 0.22999056,
0.1566043 ])
"""
# Separate the numerator and denominator for simplicity
# Calculate the numerator and denominator element-wise
numerator = np.exp(a_value * vector) - np.exp(-b_value * vector)
denominator = np.exp(c_value * vector) + np.exp(-d_value * vector)
# Calculate and return the final result element-wise
return numerator / denominator
if __name__ == "__main__":
import doctest
doctest.testmod()