Learn about modern speech recognition and the Graph Transformer Networks with @awnihannun!
In this lecture, Awni covers the connectionist temporal classification (CTC) loss, beam search decoding, weighted finite-state automata and transducers, and GTNs!
In this lecture, Awni covers the connectionist temporal classification (CTC) loss, beam search decoding, weighted finite-state automata and transducers, and GTNs!
«Graph Transformer Networks are deep learning architectures whose states are not tensors but graphs.
You can back-propagate gradients through modules whose inputs and outputs are weighted graphs.
GTNs are very convenient for end-to-end training of speech recognition and NLP sys.»
You can back-propagate gradients through modules whose inputs and outputs are weighted graphs.
GTNs are very convenient for end-to-end training of speech recognition and NLP sys.»
«They can be seen as a differentiable form of WFST (weighted finite-state transducers) widely used in speech recognition.
Awni is the lead author of libgtn, a GTN library for PyTorch.»
Awni is the lead author of libgtn, a GTN library for PyTorch.»
«GTNs are:
- Not connected with graph NN.
- Not connected with transformer architectures.
- Predates both of those by over 20 years.
- Connected with conditional random fields (think of GTNs as a kind of deep CRF), and predate those by 4 years.»
- Not connected with graph NN.
- Not connected with transformer architectures.
- Predates both of those by over 20 years.
- Connected with conditional random fields (think of GTNs as a kind of deep CRF), and predate those by 4 years.»
«GTNs are described in the second half of this paper:
Y. LeCun, L. Bottou, Y. Bengio and P. Haffner: Gradient-Based Learning Applied to Document Recognition, Proceedings of the IEEE, 86(11):2278-2324, November 1998.
yann.lecun.com/exdb/publis/pd…»
Y. LeCun, L. Bottou, Y. Bengio and P. Haffner: Gradient-Based Learning Applied to Document Recognition, Proceedings of the IEEE, 86(11):2278-2324, November 1998.
yann.lecun.com/exdb/publis/pd…»
You can read an accessible blog post by clicking on the link below.
ai.facebook.com/blog/a-new-ope…
ai.facebook.com/blog/a-new-ope…
Full detailed paper available below.
arxiv.org/abs/2010.01003
arxiv.org/abs/2010.01003
• • •
Missing some Tweet in this thread? You can try to
force a refresh
