Tf Linear regression (1차)

 

# Y = Wx+B

import numpy as np

import tensorflow as tf

print("NumPy Version :{}".format(np.__version__))

print("TensorFlow Version :{}".format(tf.__version__))

# Input(AGE) and Labels(BP)

x_input = tf.constant([25,25,25,35,35,35,45,45,45,55,55,55,65,65,65,73,73,73], dtype= tf.float32)

labels = tf.constant([118,125,130,118,126,123,120,124,130,122,125,130,127,130,130,125.5,130,138], dtype= tf.float32)

# Weights

W = tf.Variable(tf.random.normal(()), dtype=tf.float32)

B = tf.Variable(tf.random.normal(()), dtype=tf.float32)

def Hypothesis(x):

#   return x*W + B

  return tf.add(tf.multiply(x, W), B)

def Cost():

  return tf.reduce_mean(tf.square(Hypothesis(x_input)-labels))

#%%time

# Parameter Set

epochs = 150000

learning_rate = 0.0003

optimizer = tf.keras.optimizers.SGD(learning_rate=learning_rate)

# 학습 (Training)

for cnt in range(1, epochs+1):

  if cnt % (epochs//20) == 0:

    print("[{:>5}] cost = {:>10.4}, W = {:>7.4}, B = {:>7.4}".format(cnt, Cost(), W.numpy(), B.numpy()))

  optimizer.minimize(Cost,[W, B])

  # Predict

def Predict(x):

    return Hypothesis(x).numpy()

age = 50.0

print("{} Years : {:>7.4}mmHg".format(age, Predict(age)))

 

 

결과

 

[60000] cost =      18.09, W =  0.2186, B =   115.0
[67500] cost =      17.49, W =  0.1976, B =   116.2
[75000] cost =      17.24, W =  0.1842, B =   116.9
[82500] cost =      17.14, W =  0.1757, B =   117.4
[90000] cost =       17.1, W =  0.1703, B =   117.7
[97500] cost =      17.08, W =  0.1668, B =   117.9
[105000] cost =      17.08, W =  0.1647, B =   118.0
[112500] cost =      17.07, W =  0.1634, B =   118.1
[120000] cost =      17.07, W =  0.1624, B =   118.1
[127500] cost =      17.07, W =   0.162, B =   118.1
[135000] cost =      17.07, W =   0.162, B =   118.1
[142500] cost =      17.07, W =   0.162, B =   118.1
[150000] cost =      17.07, W =   0.162, B =   118.1
50.0 Years :   126.2mmHg