Research Interests Recurrent neural networks (especially LSTM) Supervised sequence labelling (especially speech and handwriting recognition) Unsupervised sequence learning
At IDSIA, he trained long-term neural memory networks by a new method called connectionist time classification. In certain applications, this method outperformed traditional voice recognition models. In 2009, his CTC-trained LSTM was the first repeat neural network to win pattern recognition contests, winning a number of handwriting awards.
This is a very popular method. Google uses CTC-trained LSTM for smartphone voice recognition.Graves also designs the neural Turing machines and the related neural computer.