如何平均多批次的摘要？

Question

假设我有一堆定义如下的摘要:

loss = ...
tf.scalar_summary("loss", loss)
# ...
summaries = tf.merge_all_summaries()

我可以summaries在训练数据的每几步评估张量,并将结果传递给a SummaryWriter.结果将是嘈杂的摘要,因为它们仅在一个批次上计算.

但是,我想计算整个验证数据集的摘要.当然,我无法将验证数据集作为单个批次传递,因为它太大了.因此,我将获得每个验证批次的摘要输出.

有没有办法对这些摘要进行平均,以便看起来好像是在整个验证集上计算了摘要？

Answer 1

在Python中对度量进行平均,并为每个均值创建一个新的Summary对象.这是我做的:

accuracies = []

# Calculate your measure over as many batches as you need
for batch in validation_set:
  accuracies.append(sess.run([training_op]))

# Take the mean of you measure
accuracy = np.mean(accuracies)

# Create a new Summary object with your measure
summary = tf.Summary()
summary.value.add(tag="%sAccuracy" % prefix, simple_value=accuracy)

# Add it to the Tensorboard summary writer
# Make sure to specify a step parameter to get nice graphs over time
summary_writer.add_summary(summary, global_step)

Answer 2

我会避免计算图表外的平均值.

您可以使用tf.train.ExponentialMovingAverage:

ema = tf.train.ExponentialMovingAverage(decay=my_decay_value, zero_debias=True)
maintain_ema_op = ema.apply(your_losses_list)

# Create an op that will update the moving averages after each training step.
with tf.control_dependencies([your_original_train_op]):
    train_op = tf.group(maintain_ema_op)

然后,使用:

sess.run(train_op)

这将调用,maintain_ema_op因为它被定义为控件依赖项.

为了获得指数移动平均线,请使用:

moving_average = ema.average(an_item_from_your_losses_list_above)

并使用以下方法检索其值:

value = sess.run(moving_average)

这会计算计算图表中的移动平均值.

Answer 3

我认为让tensorflow进行计算总是更好.

看看流媒体指标.它们具有更新功能以提供当前批次的信息和函数以获取平均摘要.看起来有点像这样:

accuracy = ... 
streaming_accuracy, streaming_accuracy_update = tf.contrib.metrics.streaming_mean(accuracy)
streaming_accuracy_scalar = tf.summary.scalar('streaming_accuracy', streaming_accuracy)

# set up your session etc. 

for i in iterations:
      for b in batches:
               sess.run([streaming_accuracy_update], feed_dict={...})

     streaming_summ = sess.run(streaming_accuracy_scalar)
     writer.add_summary(streaming_summary, i)

另请参阅tensorflow文档:https://www.tensorflow.org/versions/master/api_guides/python/contrib.metrics

这个问题: 如何在tensorflow中累积汇总统计数据