Scribd
Upload
4 views

Assignment 1

Uploaded by

ykl.zjut
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
4 views

Assignment 1

Uploaded by

ykl.zjut
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Assignment1.

py Wed Sep 12 13:42:02 2018 1


#Camille Chow
#ECE 471 Assignment 1

import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
from tqdm import tqdm
from datetime import datetime

M = 6
BATCH_SIZE = 40
NUM_BATCHES = 300

class Data(object):
def __init__(self):
num_samp = 50
sigma = 0.1
np.random.seed()

self.index = np.arange(num_samp)
self.x = np.random.uniform(size=(num_samp))
self.y = np.sin(2*np.pi*self.x) + np.random.normal(0, sigma, num_samp)

def get_batch(self):
choices = np.random.choice(self.index, size=BATCH_SIZE)

return self.x[choices], self.y[choices].flatten()

def f(x):
#to be optimized:
w = tf.get_variable(’w’, [M, 1], tf.float32, tf.random_normal_initializer())
b = tf.get_variable(’b’, [], tf.float32, tf.zeros_initializer())
mu = tf.get_variable(’mu’, [M, 1], tf.float32, tf.random_uniform_initializer())
sig = tf.get_variable(’sig’, [M, 1], tf.float32, tf.random_uniform_initializer())
#calculate phi
phi = tf.exp(- tf.pow((x - mu)/sig,2))
#calculate yhat
return tf.squeeze(tf.matmul(tf.transpose(w),phi) + b)

x = tf.placeholder(tf.float32, [BATCH_SIZE])
y = tf.placeholder(tf.float32, [BATCH_SIZE])
y_hat = f(x)

loss = .5 * tf.reduce_mean(tf.pow(y_hat - y, 2))


optim = tf.train.GradientDescentOptimizer(learning_rate=0.1).minimize(loss)
init = tf.global_variables_initializer()

sess = tf.Session()
sess.run(init)

data = Data()
#perform gradient descent
for _ in tqdm(range(0, NUM_BATCHES)):
x_np, y_np = data.get_batch()
loss_np, _ = sess.run([loss, optim], feed_dict={x: x_np, y: y_np})
#plots
plt.figure(1, figsize=[18,12])
t1 = np.linspace(0,1,BATCH_SIZE)

plt.subplot(121)

plt.plot(t1, np.sin(2 * np.pi * t1), ’k’) #perfect sine


plt.scatter(data.x, data.y) #training data
yhats = sess.run(y_hat, feed_dict={x:t1})
plt.plot(t1, yhats, ’r--’) #fit

plt.xlabel(’x’)
plt.ylabel(’y’)
plt.title(’Training Data and Fit’)
Assignment1.py Wed Sep 12 13:42:02 2018 2
plt.subplot(122)
#get mu’s and sigma’s
for var in tf.get_collection(tf.GraphKeys.TRAINABLE_VARIABLES):
if var.name.rstrip(":0") == "mu":
mus = np.array(sess.run(var)).flatten()
if var.name.rstrip(":0") == "sig":
sigs = np.array(sess.run(var)).flatten()

t2 = np.linspace(0,1)
#plot gaussians
for i in range(M):
y = np.exp(-((t2 - mus[i])/sigs[i])**2)
plt.plot(t2,y)

plt.xlabel(’x’)
plt.ylabel(’y’)
plt.title(’Bases for Fit’)

plt.show()
Training Data and Fit Bases for Fit
1.0
1.0

0.8
0.5

0.6

0.0
y

y
0.4

0.5
0.2

1.0
0.0
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
x x

You might also like