import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt


train_X = np.asarray([3.3,4.4,5.5,6.71,6.93,4.168,9.779,6.182,7.59,2.167,
                         7.042,10.791,5.313,7.997,5.654,9.27,3.1])
train_Y = np.asarray([1.7,2.76,2.09,3.19,1.694,1.573,3.366,2.596,2.53,1.221,
                         2.827,3.465,1.65,2.904,2.42,2.94,1.3])

n_samples = train_X.shape[0]


## 1. definicija računskog grafa
# podatci i parametri
X  = tf.placeholder(tf.float32, [None])
Y_ = tf.placeholder(tf.float32, [None])
a = tf.Variable(0.0)
b = tf.Variable(0.0)

# afini regresijski model
Y = a * X + b

# kvadratni gubitak
loss = (Y-Y_)**2

#derivacije kvadratnog gubitka
da = 2 * X * (a * X + b)
db = 2 * (a * X + b)

# optimizacijski postupak: gradijentni spust
trainer = tf.train.GradientDescentOptimizer(0.1)
grad = trainer.compute_gradients(loss, [a,b])
train_op = trainer.apply_gradients(grad)

## 2. inicijalizacija parametara
sess = tf.Session()
sess.run(tf.initialize_all_variables())

## 3. učenje
# neka igre počnu!
for i in range(100):
  val_loss, val_grad, val_da, val_db, _, val_a,val_b = sess.run([loss, grad, da, db, train_op, a, b],
      feed_dict={X: [1,2], Y_: [3,5]})

  print("Eksplicitno racunanje gradijenta: ", val_da, val_db)
  print("Racunanje gradijenta s compute_gradients() fun: ", val_grad[0])
  print()
  #print(i, val_grad[0],val_loss,val_a,val_b)



plt.plot([1,2], [3,5], 'ro', label='Original data')
plt.plot([1,2], sess.run(a) * [1,2] + sess.run(b), label='Fitted line')
plt.legend()
plt.show()