activation.py

activation.py

# -*- coding: utf-8 -*-
import numpy as np

class Activation():
    def __init__(self, name="sigmoid"):
        self.name = name

    def forward_prop(self, Z):
        if self.name=="sigmoid":
            return self.sigmoid(Z)
        elif self.name=="tanh":
            return self.tanh(Z)
        elif self.name=="relu":
            return self.relu(Z)
        elif self.name=="linear":
            return self.linear(Z)

    def back_prop(self, Z, dPhi):
        if self.name=="sigmoid":
            return self.grad_sigmoid(Z) * dPhi
        elif self.name=="tanh":
            return self.grad_tanh(Z) * dPhi
        elif self.name=="relu":
            return self.grad_relu(Z) * dPhi
        elif self.name=="linear":
            return self.grad_linear(Z) * dPhi

    def sigmoid(self, Z):
        return 1.0/(1.0 + np.exp(-Z))

    def grad_sigmoid(self, Z):
        return self.sigmoid(Z)*(1.0 - self.sigmoid(Z))

    def tanh(self, Z):
        return np.tanh(Z)

    def grad_tanh(self, Z):
        return 1.0 - self.tanh(Z)**2

    def relu(self, Z):
        return Z * (Z > 0)

    def grad_relu(self, Z):
        return 1.0 * (Z > 0)

    def linear(self, Z):
        return Z

    def grad_linear(self, Z):
        return np.ones(Z.shape)