在为mnist数据集构建CNN时，准确度非常低

Question

正如标题所述，我的CNN在mnist数据集上的准确度非常低（约70％）。我的架构包括两个卷积层和两个完全连接的层。我很高兴我已经开始运行，但现在我对如何优化我的代码感到困惑。任何帮助将不胜感激。

import os
import tensorflow as tf
import sys
import urllib

if sys.version_info[0] >= 3:
  from urllib.request import urlretrieve
else:
  from urllib import urlretrieve

LOGDIR = 'log3/'
GITHUB_URL ='https://raw.githubusercontent.com/mamcgrath/TensorBoard-TF-Dev-Summit-Tutorial/master/'

### MNIST EMBEDDINGS ###
mnist = tf.contrib.learn.datasets.mnist.read_data_sets(train_dir=LOGDIR + 'data', one_hot=True)
### Get a sprite and labels file for the embedding projector ###
urlretrieve(GITHUB_URL + 'labels_1024.tsv', LOGDIR + 'labels_1024.tsv')
urlretrieve(GITHUB_URL + 'sprite_1024.png', LOGDIR + 'sprite_1024.png')

# Add convolution layer
def conv_layer(input, size_in, size_out, name="conv"):
  with tf.name_scope(name):
    #w = tf.Variable(tf.zeros([5, 5, size_in, size_out]), name="W")
    #b = tf.Variable(tf.zeros([size_out]), name="B")
    w = tf.Variable(tf.truncated_normal([4, 4, size_in, size_out], stddev=0.1), name="W")
    b = tf.Variable(tf.constant(0.1, shape=[size_out]), name="B")
    conv = tf.nn.conv2d(input, w, strides=[1, 1, 1, 1], padding="SAME")
    act = tf.nn.relu(conv + b)
    tf.summary.histogram("weights", w)
    tf.summary.histogram("biases", b)
    tf.summary.histogram("activations", act)
    return tf.nn.max_pool(act, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")


# Add fully connected layer
def fc_layer(input, size_in, size_out, name="fc"):
  with tf.name_scope(name):
    w = tf.Variable(tf.truncated_normal([size_in, size_out], stddev=0.1), name="W")
    b = tf.Variable(tf.constant(0.1, shape=[size_out]), name="B")
    act = tf.nn.relu(tf.matmul(input, w) + b)
    tf.summary.histogram("weights", w)
    tf.summary.histogram("biases", b)
    tf.summary.histogram("activations", act)
    return act


def mnist_model(learning_rate, use_two_conv, use_two_fc, hparam):
  tf.reset_default_graph()
  sess = tf.Session()

  # Setup placeholders, and reshape the data
  x = tf.placeholder(tf.float32, shape=[None, 784], name="x")
  x_image = tf.reshape(x, [-1, 28, 28, 1])
  tf.summary.image('input', x_image, 3)
  y = tf.placeholder(tf.float32, shape=[None, 10], name="labels")

  if use_two_conv:
    conv1 = conv_layer(x_image, 1, 32, "conv1")
    conv_out = conv_layer(conv1, 32, 64, "conv2")
  else:
    conv1 = conv_layer(x_image, 1, 64, "conv")
    conv_out = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")

  flattened = tf.reshape(conv_out, [-1, 7 * 7 * 64])


  if use_two_fc:
    fc1 = fc_layer(flattened, 7 * 7 * 64, 1024, "fc1")
    embedding_input = fc1
    embedding_size = 1024
    logits = fc_layer(fc1, 1024, 10, "fc2")
  else:
    embedding_input = flattened
    embedding_size = 7*7*64
    logits = fc_layer(flattened, 7*7*64, 10, "fc")

  with tf.name_scope("xent"):
    xent = tf.reduce_mean(
        tf.nn.softmax_cross_entropy_with_logits(
            logits=logits, labels=y), name="xent")
    tf.summary.scalar("xent", xent)

  with tf.name_scope("train"):
    train_step = tf.train.AdamOptimizer(learning_rate).minimize(xent)

  with tf.name_scope("accuracy"):
    correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(y, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    tf.summary.scalar("accuracy", accuracy)

  summ = tf.summary.merge_all()


  embedding = tf.Variable(tf.zeros([1024, embedding_size]), name="test_embedding")
  assignment = embedding.assign(embedding_input)
  saver = tf.train.Saver()

  sess.run(tf.global_variables_initializer())
  writer = tf.summary.FileWriter(LOGDIR + hparam)
  writer.add_graph(sess.graph)

  config = tf.contrib.tensorboard.plugins.projector.ProjectorConfig()
  embedding_config = config.embeddings.add()
  embedding_config.tensor_name = embedding.name
  embedding_config.sprite.image_path = LOGDIR + 'sprite_1024.png'
  embedding_config.metadata_path = LOGDIR + 'labels_1024.tsv'
  # Specify the width and height of a single thumbnail.
  embedding_config.sprite.single_image_dim.extend([28, 28])
  tf.contrib.tensorboard.plugins.projector.visualize_embeddings(writer, config)

  for i in range(201):
    batch = mnist.train.next_batch(100)
    if i % 5 == 0:
      [train_accuracy, s] = sess.run([accuracy, summ], feed_dict={x: batch[0], y: batch[1]})
      writer.add_summary(s, i)
      print (train_accuracy)
    if i % 500 == 0:
      sess.run(assignment, feed_dict={x: mnist.test.images[:1024], y: mnist.test.labels[:1024]})
      saver.save(sess, os.path.join(LOGDIR, "model.ckpt"), i)
    sess.run(train_step, feed_dict={x: batch[0], y: batch[1]})

def make_hparam_string(learning_rate, use_two_fc, use_two_conv):
  conv_param = "conv2" if use_two_conv else "conv1"
  fc_param = "fc2" if use_two_fc else "fc1"
  return "lr_%.0E%s%s" % (learning_rate, conv_param, fc_param)

def main():
  # You can try adding some more learning rates
  #for learning_rate in [1E-3, 1E-4, 1E-5]:
  for learning_rate in [1E-4]:

    # Include "False" as a value to try different model architectures
    #for use_two_fc in [True, False]:
    for use_two_fc in [True]:
      #for use_two_conv in [True, False]:
      for use_two_conv in [True]:
        # Construct a hyperparameter string for each one (example: "lr_1E-3fc2conv2")
        hparam = make_hparam_string(learning_rate, use_two_fc, use_two_conv)
        print('Starting run for %s' % hparam)
        sys.stdout.flush() # this forces print-ed lines to show up.

        # Actually run with the new settings
        mnist_model(learning_rate, use_two_fc, use_two_conv, hparam)


if __name__ == '__main__':
  main()

更新代码：

import os
import tensorflow as tf
import sys
import urllib

if sys.version_info[0] >= 3:
  from urllib.request import urlretrieve
else:
  from urllib import urlretrieve

LOGDIR = 'log3/'
GITHUB_URL ='https://raw.githubusercontent.com/mamcgrath/TensorBoard-TF-Dev-Summit-Tutorial/master/'

### MNIST EMBEDDINGS ###
mnist = tf.contrib.learn.datasets.mnist.read_data_sets(train_dir=LOGDIR + 'data', one_hot=True)
### Get a sprite and labels file for the embedding projector ###
urlretrieve(GITHUB_URL + 'labels_1024.tsv', LOGDIR + 'labels_1024.tsv')
urlretrieve(GITHUB_URL + 'sprite_1024.png', LOGDIR + 'sprite_1024.png')

# Add convolution layer
def conv_layer(input, size_in, size_out, name="conv"):
  with tf.name_scope(name):
    #w = tf.Variable(tf.zeros([5, 5, size_in, size_out]), name="W")
    #b = tf.Variable(tf.zeros([size_out]), name="B")
    w = tf.Variable(tf.truncated_normal([4, 4, size_in, size_out], stddev=0.1), name="W")
    b = tf.Variable(tf.constant(0.1, shape=[size_out]), name="B")
    conv = tf.nn.conv2d(input, w, strides=[1, 1, 1, 1], padding="SAME")
    act = tf.nn.relu(conv + b)
    tf.summary.histogram("weights", w)
    tf.summary.histogram("biases", b)
    tf.summary.histogram("activations", act)
    return tf.nn.max_pool(act, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")


# Add fully connected layer
def fc_layer(input, size_in, size_out, name="fc"):
  with tf.name_scope(name):
    w = tf.Variable(tf.truncated_normal([size_in, size_out], stddev=0.1), name="W")
    b = tf.Variable(tf.constant(0.1, shape=[size_out]), name="B")
    act = tf.nn.relu(tf.matmul(input, w) + b)
    tf.summary.histogram("weights", w)
    tf.summary.histogram("biases", b)
    tf.summary.histogram("activations", act)
    return act


def mnist_model(learning_rate, use_two_conv, use_two_fc, hparam):
  tf.set_random_seed(1)
  tf.reset_default_graph()
  sess = tf.Session()

  # Setup placeholders, and reshape the data
  x = tf.placeholder(tf.float32, shape=[None, 784], name="x")
  x_image = tf.reshape(x, [-1, 28, 28, 1])
  tf.summary.image('input', x_image, 3)
  y = tf.placeholder(tf.float32, shape=[None, 10], name="labels")

  if use_two_conv:
    conv1 = conv_layer(x_image, 1, 32, "conv1")
    conv_out = conv_layer(conv1, 32, 64, "conv2")
  else:
    conv1 = conv_layer(x_image, 1, 64, "conv")
    conv_out = tf.nn.max_pool(conv1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding="SAME")

  flattened = tf.reshape(conv_out, [-1, 7 * 7 * 64])


  if use_two_fc:
    fc1 = fc_layer(flattened, 7 * 7 * 64, 40, "fc1")
    embedding_input = fc1
    embedding_size = 40
    logits = fc_layer(fc1, 40, 10, "fc2")
  else:
    embedding_input = flattened
    embedding_size = 40
    logits = fc_layer(flattened, 40, 10, "fc")

  with tf.name_scope("xent"):
    xent = tf.reduce_mean(
        tf.nn.softmax_cross_entropy_with_logits(
            logits=logits, labels=y), name="xent")
    tf.summary.scalar("xent", xent)

  with tf.name_scope("train"):
    train_step = tf.train.AdamOptimizer(learning_rate).minimize(xent)

  with tf.name_scope("accuracy"):
    correct_prediction = tf.equal(tf.argmax(logits, 1), tf.argmax(y, 1))
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
    tf.summary.scalar("accuracy", accuracy)

  summ = tf.summary.merge_all()


  embedding = tf.Variable(tf.zeros([100, embedding_size]), name="test_embedding")
  assignment = embedding.assign(embedding_input)
  saver = tf.train.Saver()

  sess.run(tf.global_variables_initializer())
  writer = tf.summary.FileWriter(LOGDIR + hparam)
  writer.add_graph(sess.graph)

  config = tf.contrib.tensorboard.plugins.projector.ProjectorConfig()
  embedding_config = config.embeddings.add()
  embedding_config.tensor_name = embedding.name
  embedding_config.sprite.image_path = LOGDIR + 'sprite_1024.png'
  embedding_config.metadata_path = LOGDIR + 'labels_1024.tsv'
  # Specify the width and height of a single thumbnail.
  embedding_config.sprite.single_image_dim.extend([28, 28])
  tf.contrib.tensorboard.plugins.projector.visualize_embeddings(writer, config)

  for i in range(201):
    batch = mnist.train.next_batch(150)
    if i % 5 == 0:
      [train_accuracy, s] = sess.run([accuracy, summ], feed_dict={x: batch[0], y: batch[1]})
      writer.add_summary(s, i)
      print (train_accuracy)
    #if i % 500 == 0:
      #sess.run(assignment, feed_dict={x: mnist.test.images[:1024], y: mnist.test.labels[:1024]})
      #saver.save(sess, os.path.join(LOGDIR, "model.ckpt"), i)
    sess.run(train_step, feed_dict={x: batch[0], y: batch[1]})

def make_hparam_string(learning_rate, use_two_fc, use_two_conv):
  conv_param = "conv2" if use_two_conv else "conv1"
  fc_param = "fc2" if use_two_fc else "fc1"
  return "lr_%.0E%s%s" % (learning_rate, conv_param, fc_param)

def main():
  # You can try adding some more learning rates
  #for learning_rate in [1E-3, 1E-4, 1E-5]:
  for learning_rate in [0.002]:

    # Include "False" as a value to try different model architectures
    #for use_two_fc in [True, False]:
    for use_two_fc in [True]:
      #for use_two_conv in [True, False]:
      for use_two_conv in [True]:
        # Construct a hyperparameter string for each one (example: "lr_1E-3fc2conv2")
        hparam = make_hparam_string(learning_rate, use_two_fc, use_two_conv)
        print('Starting run for %s' % hparam)
        sys.stdout.flush() # this forces print-ed lines to show up.

        # Actually run with the new settings
        mnist_model(learning_rate, use_two_fc, use_two_conv, hparam)


if __name__ == '__main__':
  main()

更新了结果：

W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.
0.106667
0.213333
0.326667
0.426667
0.466667
0.473333
0.48
0.493333
0.506667
0.506667
0.473333
0.426667
0.46
0.46
0.506667
0.433333
0.526667
0.453333
0.506667
0.513333
0.42
0.426667
0.493333
0.486667
0.48
0.493333
0.5
0.493333
0.533333
0.433333
0.5
0.426667
0.42
0.493333
0.466667
0.466667
0.433333
0.46
0.52
0.513333
0.453333

Answer 1

您的隐藏图层太大，因此很难训练网络。我将其大小从1024缩小到40，并修改了if use_two_fc语句中的块，如下所示：

if use_two_fc:
    fc1 = fc_layer(flattened, 7 * 7 * 64, 40, "fc1")
    embedding_input = fc1
    embedding_size = 40
    logits = fc_layer(fc1, 40, 10, "fc2")

我还将学习率更改为0.002。我将批量大小增加到150，所以

batch = mnist.train.next_batch(150)

我还将迭代次数增加到301.最后，为了使结果更加一致，我在行之后添加了tf.set_random_seed(1)

tf.reset_default_graph()。它将准确度提高到99％，正如您在打印的精度列表中所看到的那样：

Extracting log3/data\train-images-idx3-ubyte.gz
Extracting log3/data\train-labels-idx1-ubyte.gz
Extracting log3/data\t10k-images-idx3-ubyte.gz
Extracting log3/data\t10k-labels-idx1-ubyte.gz
Starting run for lr_2E-03conv2fc2
0.133333
0.3
0.706667
0.693333
0.686667
0.726667
0.806667
0.84
0.893333
0.826667
0.846667
0.893333
0.893333
0.9
0.92
0.946667
0.893333
0.893333
0.966667
0.946667
0.946667
0.906667
0.946667
0.96
0.946667
0.966667
0.906667
0.946667
0.96
0.946667
0.92
0.96
0.953333
0.953333
0.933333
0.9
0.953333
0.966667
0.966667
0.986667
0.966667
0.986667
0.986667
0.933333
0.973333
0.973333
0.98
0.96
0.973333
0.973333
0.973333
0.986667
0.946667
0.966667
0.926667
0.94
0.96
0.973333
0.986667
0.993333
0.993333