Výsledky vyhledávání - "Patrick Doetsch"

Returnn: The RWTH extensible training framework for universal recurrent neural networks

Autor: Albert Zeyer, Ralf Schlüter, Ilia Kulikov, Hermann Ney, Patrick Doetsch, Paul Voigtlaender

Publikováno v: ICASSP

In this work we release our extensible and easily configurable neural network training software. It provides a rich set of functional layers with a particular focus on efficient training of recurrent neural network topologies on multiple GPUs. The so

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::f55901c0e196457905e57d6caacee40b
https://doi.org/10.1109/icassp.2017.7953177

Zobrazit plný text záznamu

Inverted Alignments for End-to-End Automatic Speech Recognition

Autor: Hermann Ney, Mirko Hannemann, Ralf Schlüter, Patrick Doetsch

Publikováno v: IEEE journal of selected topics in signal processing 11(8), 1265-1273 (2017). doi:10.1109/JSTSP.2017.2752691
IEEE Journal of Selected Topics in Signal Processing

In this paper, we propose an inverted alignment approach for sequence classification systems like automatic speech recognition (ASR) that naturally incorporates discriminative, artificial-neural-network-based label distributions. Instead of aligning

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::21208efa8a2622958852a4d6ad062651

Zobrazit plný text záznamu

Bidirectional Decoder Networks for Attention-Based End-to-End Offline Handwriting Recognition

Autor: Patrick Doetsch, Albert Zeyer, Hermann Ney

Publikováno v: ICFHR

Recurrent neural networks that can be trained end-to-end on sequence learning tasks provide promising benefits over traditional recognition systems. In this paper, we demonstrate the application of an attention-based long short-term memory decoder ne

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::2bf45efe97cbefba23c36b014e82ea62
https://doi.org/10.1109/icfhr.2016.0074

Zobrazit plný text záznamu

Handwriting Recognition with Large Multidimensional Long Short-Term Memory Recurrent Neural Networks

Autor: Hermann Ney, Patrick Doetsch, Paul Voigtlaender

Publikováno v: ICFHR

Multidimensional long short-term memory recurrent neural networks achieve impressive results for handwriting recognition. However, with current CPU-based implementations, their training is very expensive and thus their capacity has so far been limite

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::87526d75b883a204bda55de95782da71
https://doi.org/10.1109/icfhr.2016.0052

Zobrazit plný text záznamu

On the Benefits of Convolutional Neural Network Combinations in Offline Handwriting Recognition

Autor: Hermann Ney, Dewi Suryani, Patrick Doetsch

Publikováno v: ICFHR

In this paper, we elaborate the advantages of combining two neural network methodologies, convolutional neural networks (CNN) and long short-term memory (LSTM) recurrent neural networks, with the framework of hybrid hidden Markov models (HMM) for rec

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::2fa44e4548c570d823dabb6c9240ad0b
https://doi.org/10.1109/icfhr.2016.0046

Zobrazit plný text záznamu

A Comprehensive Study of Deep Bidirectional LSTM RNNs for Acoustic Modeling in Speech Recognition

Autor: Hermann Ney, Albert Zeyer, Patrick Doetsch, Paul Voigtlaender, Ralf Schlüter

Publikováno v: ICASSP

We present a comprehensive study of deep bidirectional long short-term memory (LSTM) recurrent neural network (RNN) based acoustic models for automatic speech recognition (ASR). We study the effect of size and depth and train models of up to 8 layers

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::8f03ec28cbe4aefb333547378ba2721e

Zobrazit plný text záznamu

Investigation of Segmental Conditional Random Fields for large vocabulary handwriting recognition

Autor: Hermann Ney, Patrick Doetsch, Mahdi Hamdani, M. Ali Basha Shaik

Publikováno v: ICDAR

Multiple types of models are used in handwriting recognition and can be broadly categorized into generative and discriminative models. Gaussian Hidden Markov Models are used successfully in most of the systems. Discriminative training can be applied

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::c5d529d6e8e5043736e2582b21b7ce4b
https://doi.org/10.1109/icdar.2015.7333764

Zobrazit plný text záznamu

Bagging by design for continuous Handwriting Recognition using multi-objective particle swarm optimization

Autor: Hermann Ney, Patrick Doetsch, Mahdi Hamdani

Publikováno v: ICDAR

Multiple classifier systems are used to improve baseline results using different strategies. Bagging by design improves standard bagging by the minimization of intersection between the different ensembles. This work proposes the use of design bagging

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::a0c1a7036dd92fab57d65ad0fff65b2d
https://doi.org/10.1109/icdar.2015.7333763

Zobrazit plný text záznamu

Sequence-discriminative training of recurrent neural networks

Autor: Ralf Schlüter, Simon Wiesler, Hermann Ney, Patrick Doetsch, Paul Voigtlaender

Publikováno v: ICASSP

We investigate sequence-discriminative training of long shortterm memory recurrent neural networks using the maximum mutual information criterion. We show that although recurrent neural networks already make use of the whole observation sequence and

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::72ae910366ed3f65bb57344cc9bcf036
https://doi.org/10.1109/icassp.2015.7178341

Zobrazit plný text záznamu

Vyhledávací nástroje:

Upřesnit hledání