TensorFlow - 如何使用Plain TensorFlow训练ANN

Question

我是TensorFlow的新人，最近我需要使用它来训练ANN而不使用TensorFlow的高级API。

但是，以下代码出了问题：

1。定义ANN

的相关参数

n_inputs = getVectorLength(instance2Path)
n_outputs = 1
n_hidden1 = 66
n_hidden2 = 24
learning_rate = 0.01
n_epochs = 400
batch_size = 50

2。在ANN

中创建每个图层

def neuron_layer(X, n_neurons, name, activation=None):
    with tf.name_scope(name):
        n_inputs = int(X.get_shape()[1])
        stddevValue = 2 / np.sqrt(n_inputs)
        initWeight = tf.truncated_normal((n_inputs, n_neurons), stddev=stddevValue)
         W = tf.Variable(initWeight, name='weights', dtype='float')
         b = tf.Variable(tf.zeros([n_neurons]), name='biases')
         z = tf.matmul(X, W) + b
         if activation == 'relu':
             return tf.nn.relu(z)
         elif name == 'outputs':
             return tf.sigmoid(z)
         pass
     pass

3。构建TensorFlow图

 X = tf.placeholder(dtype=tf.float32, shape=(None, n_inputs), name='X')
 y = tf.placeholder(dtype=tf.float32, shape=(None), name='y')
 hidden1 = neuron_layer(X, n_hidden1, 'hidden1', activation='relu')
 hidden2 = neuron_layer(hidden1, n_hidden2, 'hidden2', activation='relu')
 outputs = neuron_layer(hidden2, n_outputs, 'outputs')
 init = tf.global_variables_initializer()

4。使用Session

运行

with tf.Session() as sess:
  print n_inputs
  sess.run(init)

  samples, labels = prepareSampleAndLabelAndFeature(ConfigVars.FeatureSelectionStrategy2, instance2Path)
  sample = np.array(samples[0])
  sample = sample.reshape(1, 170)
  sess.run(outputs, feed_dict={X: sample})
  print outputs.eval()

5。问题

运行上面的代码后，我得到以下例外：

Caused by op u'X', defined at:
  File "/Users/apple/PycharmProjects/TesTensorFlow/TrainANN/ANNTest.py", line 45, in <module>
    X = tf.placeholder(dtype=tf.float32, shape=(None, n_inputs), name='X')
  File "/Users/apple/anaconda/envs/TensorFlow_GPU/lib/python2.7/site-packages/tensorflow/python/ops/array_ops.py", line 1507, in placeholder
    name=name)
  File "/Users/apple/anaconda/envs/TensorFlow_GPU/lib/python2.7/site-packages/tensorflow/python/ops/gen_array_ops.py", line 1997, in _placeholder
    name=name)
  File "/Users/apple/anaconda/envs/TensorFlow_GPU/lib/python2.7/site-packages/tensorflow/python/framework/op_def_library.py", line 768, in apply_op
    op_def=op_def)
  File "/Users/apple/anaconda/envs/TensorFlow_GPU/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 2336, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/Users/apple/anaconda/envs/TensorFlow_GPU/lib/python2.7/site-packages/tensorflow/python/framework/ops.py", line 1228, in __init__
    self._traceback = _extract_stack()

InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'X' with dtype float
     [[Node: X = Placeholder[dtype=DT_FLOAT, shape=[], _device="/job:localhost/replica:0/task:0/cpu:0"]()]]

在我看来，问题可能依赖于X的数据类型，但我已经检查过了！ X的输入向量类似于[0,1,0,1 .....]，仅包含0-1值，标签为0和1，或典型的二元分类问题。

Answer 1

有两件事可能是错误的来源：

1. When reshaping a np.ndarray, the shape is considered as one variable.例如sample = sample.reshape([1, 170])其中[1, 170]是所需的形状。

2. 由于X被定义为大小为[none, n_inputs]的占位符，因此sample的形状必须与该形状匹配。例如n_inputs是170。