Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

How to calculate AUC with tensorflow?

I've built a binary classifier using Tensorflow and now I would like to evaluate the classifier using AUC and accuracy.

As far as accuracy is concerned, I can easily do like this:

X = tf.placeholder('float', [None, n_input])
y = tf.placeholder('float', [None, n_classes])
pred = mlp(X, weights, biases, dropout_keep_prob)
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))

When calculating AUC I use the following:

print(tf.argmax(pred, 1).dtype.name)
print(tf.argmax(pred, 1).dtype.name)

a = tf.cast(tf.argmax(pred, 1),tf.float32)
b = tf.cast(tf.argmax(y,1),tf.float32)

auc = tf.contrib.metrics.streaming_auc(a, b)

and in the training loop:

train_acc = sess.run(accuracy, feed_dict={X: batch_xs, y: batch_ys, dropout_keep_prob:1.})
train_auc = sess.run(auc, feed_dict={X: batch_xs, y: batch_ys, dropout_keep_prob:1.})

which gives me the following output (and error) error:

int64
int64
/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/array_ops.py:1197: VisibleDeprecationWarning: converting an array with ndim > 0 to an index will result in an error in the future
  result_shape.insert(dim, 1)
Net built successfully...

Starting training...

Epoch: 000/300 cost: 0.618990561
Traceback (most recent call last):
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 715, in _do_call
    return fn(*args)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 697, in _run_fn
    status, run_metadata)
  File "/usr/lib/python3.5/contextlib.py", line 66, in __exit__
    next(self.gen)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/errors.py", line 450, in raise_exception_on_not_ok_status
    pywrap_tensorflow.TF_GetCode(status))
tensorflow.python.framework.errors.FailedPreconditionError: Attempting to use uninitialized value auc/false_positives
     [[Node: auc/false_positives/read = Identity[T=DT_FLOAT, _class=["loc:@auc/false_positives"], _device="/job:localhost/replica:0/task:0/cpu:0"](auc/false_positives)]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "./mlp_.py", line 152, in <module>
    train_auc = sess.run(auc, feed_dict={X: batch_xs, y: batch_ys, dropout_keep_prob:1.})
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 372, in run
    run_metadata_ptr)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 636, in _run
    feed_dict_string, options, run_metadata)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 708, in _do_run
    target_list, options, run_metadata)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/client/session.py", line 728, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors.FailedPreconditionError: Attempting to use uninitialized value auc/false_positives
     [[Node: auc/false_positives/read = Identity[T=DT_FLOAT, _class=["loc:@auc/false_positives"], _device="/job:localhost/replica:0/task:0/cpu:0"](auc/false_positives)]]
Caused by op 'auc/false_positives/read', defined at:
  File "./mlp_.py", line 121, in <module>
    auc = tf.contrib.metrics.streaming_auc(a, b)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/metrics/python/ops/metric_ops.py", line 718, in streaming_auc
    predictions, labels, thresholds, ignore_mask)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/metrics/python/ops/metric_ops.py", line 603, in _tp_fn_tn_fp
    false_positives = _create_local('false_positives', shape=[num_thresholds])
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/contrib/metrics/python/ops/metric_ops.py", line 75, in _create_local
    collections=collections)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/variables.py", line 211, in __init__
    dtype=dtype)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/variables.py", line 319, in _init_from_args
    self._snapshot = array_ops.identity(self._variable, name="read")
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/gen_array_ops.py", line 831, in identity
    result = _op_def_lib.apply_op("Identity", input=input, name=name)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/op_def_library.py", line 704, in apply_op
    op_def=op_def)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/ops.py", line 2260, in create_op
    original_op=self._default_original_op, op_def=op_def)
  File "/usr/local/lib/python3.5/dist-packages/tensorflow/python/framework/ops.py", line 1230, in __init__
    self._traceback = _extract_stack()

I don't understand what I am doing wrong and why when using accuracy only the code runs fine but when using AUC it throws this error. Could you please hint me in the right direction to understand how to fix this?

My objective is to calculate AUC and ROC for better evaluating the binary classifier performances.

like image 505
mickkk Avatar asked Sep 11 '16 10:09

mickkk


People also ask

How to add ROC AUC as a metric in TensorFlow?

The way to add the ROC AUC as a metric on your Tensorflow / Keras project is to copy this function that computes the ROC AUC and use the function name in the model. The function only requires a little customized tf code. def auc_roc (y_true, y_pred): # can be any tensorflow metric

What is the AUC function in Python?

The auc function creates four local variables, true_positives, true_negatives, false_positives and false_negatives that are used to compute the AUC. To discretize the AUC curve, a linearly spaced set of thresholds is used to compute pairs of recall and precision values.

What is approximate AUC (area under the curve)?

Approximates the AUC (Area under the curve) of the ROC or PR curves. The AUC (Area under the curve) of the ROC (Receiver operating characteristic; default) or PR (Precision Recall) curves are quality measures of binary classifiers.

Does TensorFlow’s AUC metric support binary classification?

So, one might expect that the results will be a slightly different. NOTE Tensorflow’s AUC metric supports only binary classification. Its first argument is labels which is a Tensor whose shape matches predictions and will be cast to bool.


1 Answers

I've found the same issue on github. At the moment, it seems that you also need to run sess.run(tf.initialize_local_variables()) in order to make tf.contrib.metrics.streaming_auc() work. They're working on it.

Here you have an example demonstrating how you can solve this issue:

import tensorflow as tf

a = tf.Variable([0.1, 0.5])
b = tf.Variable([0.2, 0.6])

auc = tf.contrib.metrics.streaming_auc(a, b)

sess = tf.Session()
sess.run(tf.initialize_all_variables())
sess.run(tf.initialize_local_variables()) # try commenting this line and you'll get the error
train_auc = sess.run(auc)

print(train_auc)
like image 122
rvinas Avatar answered Oct 15 '22 12:10

rvinas