# coding: utf-8

import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
rng = np.random

#Params
learning_rate = 0.01
training_epochs = 2000
display_step = 50

#Create Training Data
n_samples = 20
train_X1 = rng.binomial(1,0.5,size=n_samples)
train_X2 = rng.binomial(1,0.3,size=n_samples)
train_X3 = rng.normal(1,1,n_samples)
W1_real = rng.normal(10,1)
W2_real = rng.normal(-3,1)
W3_real = rng.normal(10,1)
b_real = rng.normal(10,1)
rand = rng.randn(n_samples)
train_Y = W1_real*train_X1 + W2_real*train_X2 + W3_real*train_X3 + b_real + rand

#Define Graph Input
X1 = tf.placeholder("float")
X2 = tf.placeholder("float")
X3 = tf.placeholder("float")
Y = tf.placeholder("float")

W1 = tf.Variable(rng.randn(), name="weight1")
W2 = tf.Variable(rng.randn(), name="weight2")
W3 = tf.Variable(rng.randn(), name="weight3")
b = tf.Variable(rng.randn(), name="bias")

#Construct Linear Model
sum_list = [tf.mul(X1,W1),tf.mul(X2,W2),tf.mul(X3,W3)] #pred = W1*X1+W2*X2+W3*X3+b
pred_X = tf.add_n(sum_list)
pred = tf.add(pred_X,b)

#MSE
cost = tf.reduce_sum(tf.pow(pred-Y,2))/(2*n_samples)

#Gradient Descent
optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

init = tf.initialize_all_variables()

#Launch the Graph
with tf.Session() as session:
    session.run(init)

    for epoch in range(training_epochs):
        for (x1,x2,x3,y) in zip(train_X1,train_X2,train_X3,train_Y):
            session.run(optimizer, feed_dict={X1: x1, X2: x2, X3: x3, Y: y})

        #Display Logs per each steps
        if (epoch+1) % display_step==0:
            c = session.run(cost, feed_dict={X1: train_X1, X2: train_X2, X3: train_X3, Y: train_Y})
            print "Epoch:", '%04d' % (epoch+1), "cost=", "{:.9f}".format(c), \
                    "W1=",session.run(W1),"W2=",session.run(W2),"W3=",session.run(W3),"b=",session.run(b)

    print "Optimization Finished!"
    training_cost = session.run(cost,feed_dict={X1: train_X1, X2: train_X2, X3: train_X3, Y: train_Y})
    print "training_cost=",training_cost
    print "Results: predicted value (true value)"
    print "W1=",session.run(W1),"(",W1_real,")"
    print "W2=",session.run(W2),"(",W2_real,")"
    print "W3=",session.run(W3),"(",W3_real,")"
    print "b=",session.run(b),"(",b_real,")",'\n'