65 lines
1.2 KiB
Python
65 lines
1.2 KiB
Python
import autograd.numpy as np
|
|
|
|
"""
|
|
References:
|
|
https://en.wikipedia.org/wiki/Activation_function
|
|
"""
|
|
|
|
|
|
def sigmoid(z):
|
|
return 1.0 / (1.0 + np.exp(-z))
|
|
|
|
|
|
def softmax(z):
|
|
# Avoid numerical overflow by removing max
|
|
e = np.exp(z - np.amax(z, axis=1, keepdims=True))
|
|
return e / np.sum(e, axis=1, keepdims=True)
|
|
|
|
|
|
def linear(z):
|
|
return z
|
|
|
|
|
|
def softplus(z):
|
|
"""Smooth relu."""
|
|
# Avoid numerical overflow, see:
|
|
# https://docs.scipy.org/doc/numpy/reference/generated/numpy.logaddexp.html
|
|
return np.logaddexp(0.0, z)
|
|
|
|
|
|
def softsign(z):
|
|
return z / (1 + np.abs(z))
|
|
|
|
|
|
def tanh(z):
|
|
return np.tanh(z)
|
|
|
|
|
|
def relu(z):
|
|
return np.maximum(0, z)
|
|
|
|
|
|
def leakyrelu(z, a=0.01):
|
|
return np.maximum(z * a, z)
|
|
|
|
|
|
def gelu(z):
|
|
"""
|
|
Gaussian Error Linear Unit (GELU)
|
|
"""
|
|
# mainly used in transformers smoother version of relu
|
|
|
|
return 0.5 * z * (
|
|
1.0 + np.tanh(
|
|
np.sqrt(2.0 / np.pi) * (z + 0.044715 * np.power(z, 3))
|
|
)
|
|
)
|
|
|
|
|
|
def get_activation(name):
|
|
"""Return activation function by name"""
|
|
try:
|
|
return globals()[name]
|
|
except Exception:
|
|
raise ValueError("Invalid activation function.")
|