from sklearn.metrics import accuracy_score
from sklearn.tree import DecisionTreeClassifier, export_graphviz
from sklearn import tree
from sklearn.model_selection import cross_val_score, StratifiedKFold, GridSearchCV
from sklearn import preprocessing
# import seaborn as sns
import matplotlib as plt
import pylab
import re
import parse
train_set = parse.TrainingDataset('train.csv')
test_set = parse.TestDataset('test.csv')
X_index = [i for i in range(12)]
X_index = [2,4,6,7,9]
# X_index = [2,4,5,6,7,9]
Y_index = 1
train_set.read_data(X_index,Y_index)
train_set.std_data(X_index)
batch_x, batch_y = train_set.next_batch(train_set.SIZE)
train_set.reset_counter()
test_f,test_l = train_set.next_batch(100)
test_f,test_l = train_set.next_batch(100)
def train_neural_network(x):
prediction = neural_network_model(x)
# OLD VERSION:
#cost = tf.reduce_mean( tf.nn.softmax_cross_entropy_with_logits(prediction,y) )
# NEW:
cost = tf.reduce_mean(
tf.nn.softmax_cross_entropy_with_logits(logits=prediction, labels=y))
optimizer = tf.train.AdamOptimizer().minimize(cost)
# optimizer = tf.train.GradientDescentOptimizer(0.1).minimize(cost)
X_labels = [i for i in xrange(1, 785)]
train_set = parse.Training_Dataset('train.csv')
train_set.read_data(X_labels, 0, True, 10)
hm_epochs = 1000
with tf.Session() as sess:
# OLD:
#sess.run(tf.initialize_all_variables())
# NEW:
sess.run(tf.global_variables_initializer())
for epoch in range(hm_epochs):
epoch_loss = 0
train_set.reset_counter()
for _ in range(int(train_set.SIZE / batch_size)):
epoch_x, epoch_y, ix = train_set.next_batch(batch_size)
_, c = sess.run([optimizer, cost],
feed_dict={
x: epoch_x,
y: epoch_y
})
epoch_loss += c
print('Epoch', epoch, 'completed out of', hm_epochs, 'loss:',
epoch_loss)
outputt = sess.run([prediction], feed_dict={x: epoch_x})
epoch_y_ = list(epoch_y[9])
outputt_ = list(outputt[0][9])
print(epoch_y_.index(max(epoch_y_)), outputt_.index(max(outputt_)))
# print(len(epoch_y[0]))
# print len(y_),len(epoch_y),len(epoch_y[0]),(y_[7])
for it in xrange(len(epoch_y)):
y_ = list(epoch_y[it])
outputt_ = list(outputt[0][it])
# print it
print(y_.index(max(y_)), outputt_.index(max(outputt_)))
x_labels = [i for i in xrange(784)]
test_set = parse.TestDataset('test.csv')
test_set.read_data(x_labels)
f = open("output.csv", 'w')
f.write("ImageId,Label\n")
counTr = 1
print(test_set.SizE, batch_size, test_set.SizE / batch_size)
for _ in xrange(test_set.SizE / batch_size):
x_test_batch, _ = test_set.next_batch(batch_size)
y_label = sess.run([prediction], feed_dict={x: x_test_batch})
for yi in range(len(y_label[0])):
y_o = list(y_label[0][yi])
# print len(y_label[0][yi]),len(y_label)
f.write(str(counTr) + ',' + str(y_o.index(max(y_o))) + '\n')
counTr += 1
# print(counTr)
# for _ in xrange(test_set.SizE% batch_size):
x_test_batch, _ = test_set.next_batch(test_set.SizE % batch_size)
y_label = sess.run([prediction], feed_dict={x: x_test_batch})
for yi in range(len(y_label[0])):
y_o = list(y_label[0][yi])
f.write(str(counTr) + ',' + str(y_o.index(max(y_o))) + '\n')
counTr += 1
f.close()