1
$\begingroup$

I am trying to implement CTC loss in Tensorflow, but their documentation is pretty limited. So I am not sure how to approach the problem. I found a good example in Theano: https://github.com/mohammadpz/CTC-Connectionist-Temporal-Classification/blob/master/ctc_cost.py.

Are any other resources that explain the CTC loss?

Also, I am trying to understand how its forward-backward algorithm works. And what is beam decoder in the case for CTC loss?

$\endgroup$
  • $\begingroup$ Hi and welcome to this community. I removed the part regarding the implementation, given that this website focuses on answering questions related to concepts. If you want an answer to a question that involves the implementation of an algorithm, have a look at Data Science SE. Also, please, ask one question per post! $\endgroup$ – nbro Jul 1 at 19:15
1
$\begingroup$

The paper Connectionist Temporal Classification: Labeling Unsegmented Sequence Data with Recurrent Neural Networks (2006), by Alex Graves et al., mentioned in the linked TensorFlow documentation, presents and explains the CTC loss and the CTC forward-backward algorithm (in section 4.1).

Have also a look at this article Sequence Modeling With CTC (2017) by Awni Hannun.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.