trainTargetsSymbols,
trainTargetClasses,
testData,
testDataSymbols,
testTargets,
testTargetsSymbols,
validationData,
validationDataSymbols,
validationTargets,
validationTargetsSymbols,
validationTargetClasses
] = data.loadDataSet()
# Construct Graph
x = tf.placeholder(tf.float32, [None, 20])
[weights, biases, y] = nn.constructNN(x, 20, 20, [256, 256])
y_ = tf.placeholder(tf.float32, [None, 20])
# Session Setup
loss = tf.reduce_mean(tf.reduce_sum(tf.square(y_ - y), reduction_indices = [1]))
train_step = tf.train.GradientDescentOptimizer(1.0).minimize(loss)
sess = tf.Session()
init = tf.initialize_all_variables()
sess.run(init)
# Train NN
num_epochs = 5000
[
opt_weights,
for index, item in enumerate(trainTargets):
trainTargetsCombined.append(
np.concatenate((item, trainTargetsSymbols[index]), axis=0))
for index, item in enumerate(testData):
testDataCombined.append(
np.concatenate((item, testDataSymbols[index]), axis=0))
for index, item in enumerate(validationData):
validationDataCombined.append(
np.concatenate((item, validationDataSymbols[index]), axis=0))
for index, item in enumerate(validationTargets):
validationTargetsCombined.append(
np.concatenate((item, validationTargetsSymbols[index]), axis=0))
# Construct Graph
x = tf.placeholder(tf.float32, [None, 1588])
[weights, biases, y] = nn.constructNN(x, 1588, 1588, [100, 100, 100])
y_ = tf.placeholder(tf.float32, [None, 1588])
# Session Setup
loss = tf.reduce_mean(tf.reduce_sum(tf.square(y_ - y), reduction_indices=[1]))
train_step = tf.train.AdamOptimizer(0.001).minimize(loss)
sess = tf.Session()
init = tf.initialize_all_variables()
sess.run(init)
# Train NN
num_epochs = 3000
[
opt_weights, opt_biases, final_weights, final_biases, min_training_loss,
import sys, datetime
sys.dont_write_bytecode = True
import utils, data, nn
# Load Datasets
trainData = []
trainTargets = []
for a in range(6, 10):
for b in range(6, 10):
trainData.append(utils.one_hot(a) + utils.one_hot(b))
trainTargets.append(utils.one_hot((a + b) // 10) + utils.one_hot((a + b) % 10))
# Construct Graph
x = tf.placeholder(tf.float32, [None, 20])
[weights, biases, y] = nn.constructNN(x, 20, 20, [100, 100])
y_ = tf.placeholder(tf.float32, [None, 20])
# Session Setup
loss = tf.reduce_mean(tf.reduce_sum(tf.square(y_ - y), reduction_indices = [1]))
train_step = tf.train.AdamOptimizer(0.01).minimize(loss)
sess = tf.Session()
init = tf.initialize_all_variables()
sess.run(init)
# Train NN
num_epochs = 100000
[
opt_weights,