How to load sparse data with TensorFlow?

Question

There is a small snippet about loading sparse data but I have no idea how to use it.

SparseTensors don\'t play well with queues. If you use SparseTensors

Answer 1

If you are passing sparse values as inputs , you need to create sparse placeholders using tf.sparse_placeholder.

You should then convert your sparse tensors to dense tensor using tf.sparse_to_dense.

For this you need to explicitly pass the sparse matrix's values , shape and indices while feeding the data in feed_dict and then later use tf.sparse_to_dense in the graph.

In the graph :

dense = tf.sparse_to_dense(
    sparse_indices=sparse_placeholder.indices,
    output_shape=sparse_placeholder.shape,
    sparse_values=sparse_placeholder.values,
validate_indices=False)

In the feed_dict:

sparse_placeholder:tf.SparseTensorValue(indices=indices,values=sparse_values,dense_shape=sparse_shape)