lab_10_5 dropout about deeplearningzerotoall

[kkweon]

The keep_prob is a hyperparameter, meaning you have to find the best value and store into that variable.

tf.placeholder is simply a variable.
There is no such thing as 0.7 being the default parameter.
Later in the same file, you can see we're setting 0.7 to that placeholder in fact.

    for i in range(total_batch):
        batch_xs, batch_ys = mnist.train.next_batch(batch_size)

        # Set keep_prob to 0.7
        feed_dict = {X: batch_xs, Y: batch_ys, keep_prob: 0.7}

        c, _ = sess.run([cost, optimizer], feed_dict=feed_dict)
        avg_cost += c / total_batch

[cjr0106]

Thanks so much ,i got it~ 2018-07-06 01:42:58"Mo Kweon" <[email protected]>写道： The keep_prob is a hyperparameter, meaning you have to find the best value and store into that variable. tf.placeholder is simply a variable. There is no such thing as 0.7 being the default parameter. Later in the same file, you can see we're setting 0.7 to that placeholder in fact. for i inrange(total_batch): batch_xs, batch_ys = mnist.train.next_batch(batch_size) # Set keep_prob to 0.7 feed_dict = {X: batch_xs, Y: batch_ys, keep_prob: 0.7} c, _ = sess.run([cost, optimizer], feed_dict=feed_dict) avg_cost += c / total_batch — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[github-actions]

Stale issue message