I am using Tensorflow's tf.nn.ctc_beam_search_decoder() to decode the output of a RNN doing some many-to-many mapping (i.e., multiple softmax outputs for each network cell).
A simplified version of the network's output and the Beam search decoder is:
import numpy as np
import tensorflow as tf
batch_size = 4
sequence_max_len = 5
num_classes = 3
y_pred = tf.placeholder(tf.float32, shape=(batch_size, sequence_max_len, num_classes))
y_pred_transposed = tf.transpose(y_pred,
perm=[1, 0, 2]) # TF expects dimensions [max_time, batch_size, num_classes]
logits = tf.log(y_pred_transposed)
sequence_lengths = tf.to_int32(tf.fill([batch_size], sequence_max_len))
decoded, log_probabilities = tf.nn.ctc_beam_search_decoder(logits,
sequence_length=sequence_lengths,
beam_width=3,
merge_repeated=False, top_paths=1)
decoded = decoded[0]
decoded_paths = tf.sparse_tensor_to_dense(decoded) # Shape: [batch_size, max_sequence_len]
with tf.Session() as session:
tf.global_variables_initializer().run()
softmax_outputs = np.array([[[0.1, 0.1, 0.8], [0.8, 0.1, 0.1], [0.8, 0.1, 0.1], [0.8, 0.1, 0.1], [0.8, 0.1, 0.1]],
[[0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7]],
[[0.1, 0.7, 0.2], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7]],
[[0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7], [0.1, 0.2, 0.7]]])
decoded_paths = session.run(decoded_paths, feed_dict = {y_pred: softmax_outputs})
print(decoded_paths)
The output in this case is:
[[0]
[1]
[1]
[1]]
My understanding is that the output tensor should be of dimensions [batch_size, max_sequence_len], with each row containing the indices of the relevant classes in the found path.
In this case I would expect the output to be similar to:
[[2, 0, 0, 0, 0],
[2, 2, 2, 2, 2],
[1, 2, 2, 2, 2],
[2, 2, 2, 2, 2]]
What am I not understanding about how ctc_beam_search_decoder works?