Deep Learning By Example

Placeholders

The next type of nodes are placeholders. Placeholders are nodes whose values are fed at execution time:

If you have inputs into your computational graph that depend on some external data, these are placeholders for values that we are going to add into our computation during training. So, for placeholders, we don't provide any initial values. We just assign a data type and shape of a tensor so the graph still knows what to compute even though it doesn't have any stored values yet.

We can use the placeholder function of TensorFlow to create a placeholder:

ph_var1 = tf.placeholder(tf.float32,shape=(2,3))
ph_var2 = tf.placeholder ...