rnn : Recurrent Library for Torch

By Nicholas Léonard, Sagar Waghmare, Yang Wang, Jin-Hwa Kim. 2015

43 citations recorded by the source corpus.

The rnn package provides components for implementing a wide range of Recurrent Neural Networks. It is built withing the framework of the Torch distribution for use with the nn package. The components have evolved from 3 iterations, each adding to the flexibility and capability of the package. All component modules inherit either the AbstractRecurrent or AbstractSequencer classes. Strong unit testing, continued backwards compatibility and access to supporting material are the principles followed during its development. The package is compared against existing implementations of two published papers.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures, ML Systems & Hardware, Signal Processing

How researchers see this paper

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

In Artificial Intelligence

1. Mohit Bansal
2. Graham Neubig
3. Alexander M. Rush
4. Richard Socher
5. Caiming Xiong
6. Sebastian Gehrmann
7. Jianfeng Gao
8. Luke S. Zettlemoyer

In Computer Vision and Pattern Recognition

1. Dacheng Tao
2. Luc Van Gool
3. Xiaogang Wang
4. Ming-Hsuan Yang
5. Trevor Darrell
6. Jitendra Malik
7. Andrew Zisserman
8. Yi Yang

In Model Architectures

1. Alexander M. Rush
2. Yarin Gal
3. Roberto Cipolla
4. Sebastian Gehrmann
5. Karen Livescu
6. Tomás Kociský
7. Yizhe Zhang
8. Sam Wiseman

In ML Systems & Hardware

1. Alexander M. Rush
2. Yarin Gal
3. Karen Livescu
4. Shane Settle
5. Bartlomiej Romanski
6. Konrad Zolna
7. Lin F. Yang
8. Phil Blunsom

In Signal Processing

1. Shane Settle
2. Karen Livescu
3. Sebastian Gehrmann
4. Hendrik Strobelt
5. Alexander M. Rush
6. Yarin Gal
7. Bartlomiej Romanski
8. Konrad Zolna

Citations in the corpus

• Spatio-temporal video autoencoder with differentiable memory (2015, methodology): “Our implementation is based on Torch library Collobert et al. and extends the rnn package L´eonard et al. (2015).”
• Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks (2016, methodology): “For our experiments we utilized the Torch framework and the Element RNN library [16].”
• Discriminative acoustic word embeddings: Tecurrent neural network-based approaches (2016, methodology): “All models were implemented in Torch [38] and used the rnn library of [39].”
• Learning Global Features for Coreference Resolution (2016, methodology): “We use a single-layer LSTM (without “peep-hole” connections), as implemented in the element-rnn library (Léonard et al., 2015).”
• Combining Fully Convolutional and Recurrent Neural Networks for 3D Biomedical Image Segmentation (2016, methodology): “Our framework was implemented in Torch7 [5] and the RNN package [12].”
• Uncertainty in Deep Learning (2016, result): “Our language modelling results were further reproduced in the work of [Léonard et al., 2015] – an open source Torch implementation of RNNs, which includes our dropout variant for RNNs.”
• User Modeling Using LSTM Networks (2017, methodology): “Our LSTM model was trained using the rnn Torch library (Léonard et al. 2015).”
• Multimodal Residual Learning for Visual QA (2016, methodology): “Torch framework and rnn package [14] are used to build our models.”
• Optimizing Performance of Recurrent Neural Networks on GPUs (2016, methodology): “While GPUs are already widely used to compute RNNs [15, 16, 17, 18], there has been less work on the optimization of RNN runtime.”