In [1]:

    

import os
import numpy as np
import tensorflow as tf
import matplotlib.pyplot as plt
print ("Packages loaded")


    

    




Packages loaded

In [2]:

    

# Load them!
cwd = os.getcwd()
loadpath = cwd + "/data/custom_data.npz"
l = np.load(loadpath)

# See what's in here
print (l.files)

# Parse data
trainimg = l['trainimg']
trainlabel = l['trainlabel']
testimg = l['testimg']
testlabel = l['testlabel']
use_gray = l['use_gray']
ntrain = trainimg.shape[0]
nclass = trainlabel.shape[1]
dim    = trainimg.shape[1]
ntest  = testimg.shape[0]
print ("%d train images loaded" % (ntrain))
print ("%d test images loaded" % (ntest))
print ("%d dimensional input" % (dim))
print ("%d classes" % (nclass))


    

    




['trainlabel', 'imgsize', 'trainimg', 'testimg', 'testlabel', 'use_gray']
52 train images loaded
35 test images loaded
4096 dimensional input
2 classes

In [3]:

    

tf.set_random_seed(0)
# Parameters of Logistic Regression
learning_rate   = 0.001
training_epochs = 1000
batch_size      = 10
display_step    = 100

# Create Graph for Logistic Regression
x = tf.placeholder("float", [None, dim]) 
y = tf.placeholder("float", [None, nclass]) 
W = tf.Variable(tf.zeros([dim, nclass]), name = 'weights')
b = tf.Variable(tf.zeros([nclass]))


    

In [4]:

    

WEIGHT_DECAY_FACTOR = 1 #  0.000001
l2_loss = tf.add_n([tf.nn.l2_loss(v) 
            for v in tf.trainable_variables()])
_pred = tf.nn.softmax(tf.matmul(x, W) + b) 
cost = tf.reduce_mean(-tf.reduce_sum(y*tf.log(_pred)
                                     , reduction_indices=1)) 
cost = cost + WEIGHT_DECAY_FACTOR*l2_loss
optm = tf.train.GradientDescentOptimizer(
    learning_rate).minimize(cost) 
_corr = tf.equal(tf.argmax(_pred, 1), tf.argmax(y, 1))    
accr = tf.reduce_mean(tf.cast(_corr, tf.float32))
init = tf.initialize_all_variables()
print ("Functions ready")


    

    




Functions ready

In [5]:

    

# Launch the graph
sess = tf.Session()
sess.run(init)
# Training cycle
for epoch in range(training_epochs):
    avg_cost = 0.
    num_batch = int(ntrain/batch_size)
    # Loop over all batches
    for i in range(num_batch): 
        randidx = np.random.randint(ntrain, size=batch_size)
        batch_xs = trainimg[randidx, :]
        batch_ys = trainlabel[randidx, :]                
        # Fit training using batch data
        sess.run(optm, feed_dict={x: batch_xs, y: batch_ys})
        # Compute average loss
        avg_cost += sess.run(cost
            , feed_dict={x: batch_xs, y: batch_ys})/num_batch

    # Display logs per epoch step
    if epoch % display_step == 0:
        print ("Epoch: %03d/%03d cost: %.9f" % 
               (epoch, training_epochs, avg_cost))
        train_acc = sess.run(accr, feed_dict={x: batch_xs, y: batch_ys})
        print (" Training accuracy: %.3f" % (train_acc))
        test_acc = sess.run(accr, feed_dict={x: testimg, y: testlabel})
        print (" Test accuracy: %.3f" % (test_acc))
print ("Optimization Finished!")


    

    




Epoch: 000/1000 cost: 0.626751423
 Training accuracy: 0.600
 Test accuracy: 0.686
Epoch: 100/1000 cost: 0.438359487
 Training accuracy: 0.800
 Test accuracy: 0.714
Epoch: 200/1000 cost: 0.367508155
 Training accuracy: 0.900
 Test accuracy: 0.686
Epoch: 300/1000 cost: 0.363990796
 Training accuracy: 1.000
 Test accuracy: 0.714
Epoch: 400/1000 cost: 0.406193763
 Training accuracy: 0.900
 Test accuracy: 0.714
Epoch: 500/1000 cost: 0.400928861
 Training accuracy: 0.900
 Test accuracy: 0.714
Epoch: 600/1000 cost: 0.366956294
 Training accuracy: 0.900
 Test accuracy: 0.686
Epoch: 700/1000 cost: 0.334192312
 Training accuracy: 1.000
 Test accuracy: 0.686
Epoch: 800/1000 cost: 0.392353541
 Training accuracy: 1.000
 Test accuracy: 0.714
Epoch: 900/1000 cost: 0.383330226
 Training accuracy: 1.000
 Test accuracy: 0.714
Optimization Finished!

In [6]:

    

sess.close()
print ("Session closed.")


    

    




Session closed.