Blocks and Fuel: Frameworks for deep learning

Merriënboer, Bart van; Bahdanau, Dzmitry; Dumoulin, Vincent; Serdyuk, Dmitriy; Warde-Farley, David; Chorowski, Jan; Bengio, Yoshua

doi:10.48550/arxiv.1506.00619

Cited by 23 publications

(27 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The BASIC NMT system is built using the Blocks framework (van Merriënboer et al, 2015) based on the Theano library (Bastien et al, 2012) with standard hyper-parameters : the encoder and decoder networks consist of 1000 gated recurrent units (Cho et al, 2014). The decoder uses a single maxout (Goodfellow et al, 2013) output layer with the feed-forward attention model .…”

Section: Experimental Evaluationmentioning

confidence: 99%

Syntactically Guided Neural Machine Translation

Stahlberg

Hasler

Waite

et al. 2016

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)

View full text Add to dashboard Cite

We investigate the use of hierarchical phrase-based SMT lattices in end-to-end neural machine translation (NMT). Weight pushing transforms the Hiero scores for complete translation hypotheses, with the full translation grammar score and full ngram language model score, into posteriors compatible with NMT predictive probabilities. With a slightly modified NMT beam-search decoder we find gains over both Hiero and NMT decoding alone, with practical advantages in extending NMT to very large input and output vocabularies.

show abstract

Section: Experimental Evaluationmentioning

confidence: 99%

Syntactically Guided Neural Machine Translation

Stahlberg

Hasler

Waite

et al. 2016

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers)

View full text Add to dashboard Cite

show abstract

“…It is free, supported by many commercial and non-commercial software platforms and programming languages (MATLAB, C, C++, Fortran, Java), and has efficient partial I/O access to data on disk. Not surprisingly, it has become of wide use in public benchmark data sets and frameworks in fields like deep learning [10].…”

Section: A the Hdf5 Formatmentioning

confidence: 99%

VeHIF: An Accessible Vegetation High-Impedance Fault Data Set Format

Gomes¹,

Ozansoy²

2021

Preprint

View full text Add to dashboard Cite

High-impedance faults are a challenging problem in power distribution systems. They often do not trigger protection devices and can result in serious hazards such as igniting fires when in contact with vegetation. The current research field dedicated to studying these faults is extensive but suffers from a constraining bottleneck of a lack of real experimental data. Many works set to detect and localize such faults rely on highimpedance fault low-fidelity models, and the lack of public data sets makes it impractical to have objective performance benchmarks. This letter describes and proposes a format for a data set of more than 900 vegetation high-impedance faults funded by the Victorian Government in Australia recorded in high-sampling resolution. The original data set is public, but it was made available through an obscure format that limits its accessibility. The presented format in this letter uses the standard hierarchical data format (HDF5), which makes it easily accessible in many languages such as MATLAB, Python, C++, and more. The data set compiler and visualizer script are also provided in the work repository 1 .

show abstract

“…The selected model used on the test set is the one with best [23] based on Theano [24], [25]. Results on the test set using the baseline model and using the different types of convolutional layers are reported in Table I.…”

Section: Experiments a Experiments On Emotion Classification: Fau-aib...mentioning

confidence: 99%

Convolutional RNN: An enhanced model for extracting features from sequential data

Keren

Schuller

2016

2016 International Joint Conference on Neural Networks (IJCNN)

115

View full text Add to dashboard Cite

Traditional convolutional layers extract features from patches of data by applying a non-linearity on an affine function of the input. We propose a model that enhances this feature extraction process for the case of sequential data, by feeding patches of the data into a recurrent neural network and using the outputs or hidden states of the recurrent units to compute the extracted features. By doing so, we exploit the fact that a window containing a few frames of the sequential data is a sequence itself and this additional structure might encapsulate valuable information. In addition, we allow for more steps of computation in the feature extraction process, which is potentially beneficial as an affine function followed by a non-linearity can result in too simple features. Using our convolutional recurrent layers, we obtain an improvement in performance in two audio classification tasks, compared to traditional convolutional layers. Tensorflow code for the convolutional recurrent layers is publicly available in https://github.com/cruvadom/Convolutional-RNN.

show abstract

Blocks and Fuel: Frameworks for deep learning

Cited by 23 publications

References 3 publications

Syntactically Guided Neural Machine Translation

Syntactically Guided Neural Machine Translation

VeHIF: An Accessible Vegetation High-Impedance Fault Data Set Format

Convolutional RNN: An enhanced model for extracting features from sequential data

Contact Info

Product

Resources

About