Created
March 19, 2016 15:57
-
-
Save harujoh/65b2568acb011a8a9128 to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import numpy as np | |
class MLP (object): | |
""" | |
3 Layered Perceptron | |
""" | |
def __init__( self, n_input_units , n_hidden_units, n_output_units): | |
self.nin = n_input_units | |
self.nhid = n_hidden_units | |
self.nout = n_output_units | |
self.v = np .random.uniform(-1.0, 1.0, ( self.nhid, self .nin+1)) | |
self.w = np .random.uniform(-1.0, 1.0, ( self.nout, self .nhid+1)) | |
def fit( self, inputs , targets, learning_rate=0.8, epochs =10000): | |
inputs = self .__add_bias(inputs, axis=1) | |
targets = np .array(targets) | |
for loop_cnt in range(epochs): | |
# randomise the order of the inputs | |
p = np.random .randint(inputs.shape[0]) | |
xp = inputs[p] | |
bkp = targets[p] | |
# forward phase | |
gjp = self.__sigmoid(np .dot(self.v, xp)) | |
gjp = self.__add_bias(gjp) | |
gkp = self.__sigmoid(np .dot(self.w, gjp)) | |
# backward phase(back prop) | |
eps2 = self.__sigmoid_deriv(gkp) * (gkp - bkp) | |
eps = self.__sigmoid_deriv(gjp) * np .dot(self.w.T, eps2) | |
gjp = np.atleast_2d(gjp) | |
eps2 = np.atleast_2d(eps2) | |
self.w = self .w - learning_rate * np.dot(eps2.T, gjp) | |
xp = np.atleast_2d(xp) | |
eps = np.atleast_2d(eps) | |
self.v = self .v - learning_rate * np.dot(eps.T, xp)[1:, :] | |
def __add_bias( self, x , axis=None): | |
return np .insert(x, 0, 1, axis= axis) | |
def __sigmoid( self, u ): | |
""" Sigmoid function(Activation function) """ | |
return (1.0 / (1.0 + np .exp(-u))) | |
def __sigmoid_deriv( self, u ): | |
return (u * (1 - u)) | |
def predict( self, i ): | |
i=self .__add_bias(i) | |
gjp = self.__sigmoid(np .dot(self.v, i)) | |
gjp = self.__add_bias(gjp) | |
return self .__sigmoid(np.dot( self.w, gjp)) | |
# initialize | |
mlp = MLP(n_input_units=2, n_hidden_units=2, n_output_units=1) | |
inputs = np.array([[0, 0], [0, 1], [1, 0], [1, 1]]) | |
targets = np.array([0, 1, 1, 0]) | |
# training | |
mlp.fit(inputs, targets) | |
# predict | |
print ("--- predict ---") | |
for i in [[0, 0], [0, 1], [1, 0], [1, 1]]: | |
print (i, mlp.predict(i)) |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment