Kategoriale Variablen in TensorFlow

Question

Ich versuche, TensorFlow für einen Datensatz mit einigen kategorialen Variablen zu verwenden. Ich habe sie mit Dummies kodiert, aber es sieht so aus, als ob es Probleme verursacht und TF beschwert sich, dass der Datensatz nicht dicht ist.

Oder ist der Grund für den Fehler etwas völlig anderes?

Ich versuche, ein einfaches neuronales Netzwerkmodell mit 1 versteckter Schicht mit stochastischem Gradienten auszuführen. Der Code wurde zu arbeiten, wenn der Eingang numerischer Variablen (Bilder von Ziffern von MNIST)

dank

-------------------------------------------------------------------------- ValueError Traceback (most recent call last) in() 37 return(test_acc,round(l,5)) 38 ---> 39 define_batch(0.005) 40 run_batch()

in define_batch(beta) 11 shape=(batch_size, num_var)) 12 tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels)) ---> 13 tf_valid_dataset = tf.constant(valid_dataset) 14 tf_test_dataset = tf.constant(test_dataset) 15

/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/ops/constant_op.pyc in constant(value, dtype, shape, name) 159 tensor_value = attr_value_pb2.AttrValue() 160 tensor_value.tensor.CopyFrom( --> 161 tensor_util.make_tensor_proto(value, dtype=dtype, shape=shape)) 162 dtype_value = attr_value_pb2.AttrValue(type=tensor_value.tensor.dtype) 163 const_tensor = g.create_op(

/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/tensorflow/python/framework/tensor_util.pyc in make_tensor_proto(values, dtype, shape) 320 nparray = np.array(values, dtype=np_dt) 321 if list(nparray.shape) != _GetDenseDimensions(values): --> 322 raise ValueError("Argument must be a dense tensor: %s" % values) 323 # python/numpy default float type is float64. We prefer float32 instead. 324 if (nparray.dtype == np.float64) and dtype is None:

ValueError: Argument must be a dense tensor: Tuesday Wednesday Thursday Friday Saturday Sunday CENTRAL \ 736114
0.0 0.0 0.0 0.0 1.0 0.0 0.0 437148 0.0 0.0 1.0 0.0 0.0 0.0 0.0 605041 0.0 0.0 0.0 0.0 0.0 0.0 0.0 444608 0.0 0.0 0.0 0.0 1.0 0.0 0.0 695549 0.0 0.0 0.0 0.0 1.0 0.0 0.0 662807 0.0 0.0 0.0 1.0 0.0 0.0 0.0 238635 0.0 0.0 0.0 0.0 0.0 1.0 0.0 549524 0.0 0.0 0.0 1.0 0.0 0.0 0.0 705478 1.0 0.0 0.0 0.0 0.0 0.0 0.0 557716 0.0 0.0 0.0 1.0 0.0 0.0 0.0 41808 0.0 0.0 0.0 0.0 0.0 1.0 0.0 227235 1.0 0.0 0.0 0.0 0.0 0.0 0.0 848719 0.0 0.0 0.0 0.0 0.0 0.0 0.0 731202 0.0 0.0 0.0 0.0 1.0 0.0 0.0 467516 1.0 0.0 0.0 0.0 0.0 0.0 1.0

Hier ist ein Auszug aus dem Code

# Adding regularization to the 1 hidden layer network 
def define_batch(beta): 
    batch_size = 128 
    num_RELU =256 
    graph1 = tf.Graph() 
    with graph1.as_default(): 

     # Input data. For the training data, we use a placeholder that will be fed 
     # at run time with a training minibatch. 
     tf_train_dataset = tf.placeholder(tf.float32, 
             shape=(batch_size, num_var)) 
     tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels)) 
     tf_valid_dataset = tf.constant(valid_dataset) 
     tf_test_dataset = tf.constant(test_dataset) 

     # Variables. 
     weights_RELU = tf.Variable(
     tf.truncated_normal([num_var, num_RELU])) 
     biases_RELU = tf.Variable(tf.zeros([num_RELU])) 
     weights_layer1 = tf.Variable(
     tf.truncated_normal([num_RELU, num_labels])) 
     biases_layer1 = tf.Variable(tf.zeros([num_labels])) 

     # Training computation. 
     logits_RELU = tf.matmul(tf_train_dataset, weights_RELU) + biases_RELU 
     RELU_vec = tf.nn.relu(logits_RELU) 
     logits_layer = tf.matmul(RELU_vec, weights_layer1) + biases_layer1     
     # loss = tf.reduce_mean(
     #  tf.nn.softmax_cross_entropy_with_logits(logits_layer, tf_train_labels)) 
     cross_entropy = tf.nn.softmax_cross_entropy_with_logits(logits_layer, tf_train_labels,name="cross_entropy") 
     l2reg = tf.reduce_sum(tf.square(weights_RELU))+tf.reduce_sum(tf.square(weights_layer1)) 
     beta = 0.005 
     loss = tf.reduce_mean(cross_entropy+beta*l2reg) 

    # Optimizer. 
     optimizer = tf.train.GradientDescentOptimizer(0.3).minimize(loss) 

     # Predictions for the training, validation, and test data. 
     train_prediction = tf.nn.softmax(logits_layer) 
     valid_prediction = tf.nn.softmax(
     tf.matmul(tf.nn.relu((tf.matmul(tf_valid_dataset, weights_RELU) + biases_RELU)),weights_layer1)+biases_layer1) 

     test_prediction =tf.nn.softmax(
     tf.matmul(tf.nn.relu((tf.matmul(tf_test_dataset, weights_RELU) + biases_RELU)),weights_layer1)+biases_layer1) 

import datetime 

startTime = datetime.datetime.now() 

num_steps = 301 # change to 3001 

def run_batch(): 

    with tf.Session(graph=graph1) as session: 

     tf.initialize_all_variables().run() 
     print("Initialized") 
     for step in range(num_steps): 
     # Pick an offset within the training data, which has been randomized. 
     # Note: we could use better randomization across epochs. 
     offset = (step * batch_size) % (train_labels.shape[0] - batch_size) 
     # Generate a minibatch. 
     batch_data = train_dataset[offset:(offset + batch_size), :] 
     batch_labels = train_labels[offset:(offset + batch_size), :] 
     # Prepare a dictionary telling the session where to feed the minibatch. 
     # The key of the dictionary is the placeholder node of the graph to be fed, 
     # and the value is the numpy array to feed to it. 
     feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels} 
     _, l, predictions, logits = session.run(
      [optimizer, loss,train_prediction,logits_RELU], feed_dict=feed_dict) 
     if (step % 500 == 0): 
      print("Minibatch loss at step %d: %f" % (step, l)) 
      print("Minibatch accuracy: %.1f%%" % accuracy(predictions, batch_labels)) 
      print("Validation accuracy: %.1f%%" % accuracy(
      valid_prediction.eval(), valid_labels)) 
     test_acc = accuracy(test_prediction.eval(), test_labels) 
     print("Test accuracy: %.1f%%" % test_acc) 

     print('loss=%s' % l) 
    x = datetime.datetime.now() - startTime 
    print(x) 
    return(test_acc,round(l,5)) 

define_batch(0.005) 
run_batch() 

EDIT war: @gdhal Dank für es

train_dataset suchen, ist ein Pandas Datenrahmen

train_dataset.columns 
Index([u'Tuesday', u'Wednesday', u'Thursday', u'Friday', u'Saturday', 
     u'Sunday', u'CENTRAL', u'INGLESIDE', u'MISSION', u'NORTHERN', u'PARK', 
     u'RICHMOND', u'SOUTHERN', u'TARAVAL', u'TENDERLOIN', u' 3H - 4H', 
     u' 5H - 6H', u' 7H - 8H', u' 9H - 10H', u'11H - 12H', u'13H - 14H', 
     u'15H - 16H', u'17H - 18H', u'19H - 20H', u'21H - 22H', u'23H - 0H', 
     u'Xnorm', u'Ynorm', u'Hournorm'], 
     dtype='object') 

Alle Variablen sind Dummies (mit 0 oder 1 Werten) mit Ausnahme der letzten 3 Variablen (Xnorm, Ynorm und Hournorm), bei denen es sich um auf das [0,1] Intervall normierte numerische Werte handelt. valid_datasettest_dataset und haben das gleiche Format

train_labels eine Serie pandas ist

train_labels.describe() 

count   790184 
unique    39 
top  LARCENY/THEFT 
freq    157434 
Name: Category, dtype: object 

valid_labels und test_labels haben das gleiche Format

Answer 1

Versuchen eine numpy Array anstelle eines pandas Datenrahmen zugeführt wird.