Forward-backward retraining of recurrent neural networks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Speech recognition with deep recurrent neural networks (2013, methodology): “Similar procedures have been used elsewhere in speech and handwriting recognition to integrate out over possible seg-mentations [18, 19]; however CTC differs in that it ignores segmentation altogether and sums over single-timestep label decisions instead.”
Fast and accurate recurrent neural network acoustic models for speech recognition (2015, background): “While conventional hybrid speech and handwriting recognition systems usually train from fixed alignments, the use of the forward-backward algorithm to reestimate network targets given the current model can equally be applied to conventional recurrent [17] or feed-forward networks [18] if no such ali”
Supervised Sequence Labelling with Recurrent Neural Networks (2012, background): “Derived from feedforward neural networks, RNNs can use their internal state (memory) to process variable length sequences of inputs.”
On supervised learning from sequential data with applications for speech regognition (1999, methodology): “, 1996), giving a distinct best state sequence for every training sequence, although training based on a soft alignment is also possible (Senior and Robinson, 1996).”