A comparative study of open source deep learning frameworks

Shatnawi, Anas; Al-Bdour, Ghadeer; Al-Qurran, Raffi; Al‐Ayyoub, Mahmoud

doi:10.1109/iacs.2018.8355444

Cited by 49 publications

(25 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Written from another deep learning library called DistBelief, TensorFlow is implemented based on directed graphs. In these graphs, nodes represent mathematics, operations and edges represents the flow of data among the nodes, what makes TensorFlow used in any domain that the computation can be designed as a flow network (Parvat et al, 2017;Shatnawi et al, 2018).…”

Section: Tensorflowmentioning

confidence: 99%

“…The deep learning idea is to train an Artificial Neural Network (ANN) of multiple layers in a set of data in order to allow it to deal with real world tasks. Although the theoretical concepts behind are not new, deep learning has become to be a trend in the last decade due to many factors, including its well-succeed application in a variety of problem solution (many of them are potentially commercial, such as the development of new computer architectures with a higher level of parallelism, the design of Convolutional Neural Network (CNN) and a higher accessibility to high performance computers (Shatnawi et al, 2018;Verhelst and Moons, 2017) (Raschka, 2015).…”

Section: Introductionmentioning

confidence: 99%

“…Each one of these libraries has different features of performance and applies different techniques to optimize the implementation of algorithms. Therefore, even though implementing the same algorithm in different structures, the performance of these different implementations may have a considerable variation (Bahrampour et al, 2015;Shatnawi et al, 2018).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Performance Analysis of Deep Learning Libraries: TensorFlow and PyTorch

Florencio¹,

Valenç²,

Moreno³

et al. 2019

Journal of Computer Science

View full text Add to dashboard Cite

Through the increase in deep learning study and use, in the last years there was a development of specific libraries for Deep Neural Network (DNN). Each one of these libraries has different performance results and applies different techniques to optimize the implementation of algorithms. Therefore, even though implementing the same algorithm and using different libraries, the performance of their executions may have a considerable variation. For this reason, developers and scientists that work with deep learning need scientific experimental studies that examine the performance of those libraries. Therefore, this paper has the aim of evaluating and comparing these two libraries: TensorFlow and PyTorch. We have used three parameters: Hardware utilization, hardware temperature and execution time in a context of heterogeneous platforms with CPU and GPU. We used the MNIST database for training and testing the LeNet Convolutional Neural Network (CNN). We performed a scientific experiment following the Goal Question Metrics (GQM) methodology, data were validated through statistical tests. After data analysis, we show that PyTorch library presented a better performance, even though the TensorFlow library presented a greater GPU utilization rate.

show abstract

Section: Tensorflowmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Performance Analysis of Deep Learning Libraries: TensorFlow and PyTorch

Florencio¹,

Valenç²,

Moreno³

et al. 2019

Journal of Computer Science

View full text Add to dashboard Cite

show abstract

“…CNTK is a deep learning framework that givesflexibility, efficiency, balancesand Performance in one package. The inspiration behind CNTK was Lego Bricks in which, each brick is very simple and performs a specific function but by combining many bricks an arbitrary objectcan be created [2].…”

Section: Cntkmentioning

confidence: 99%

Review: Comparative Analysis of Different Techniques of DL-Frameworks

K.¹,

Jigar²,

Vyoma³

2018

IJCA

View full text Add to dashboard Cite

Deep learning is developed in 2006 and it is a part of machine learning and also for the artificial intelligence. Deep learning provides state of the arts solutions of various problems in areas like speech recognition and processing, neural language processing, image processing, computer vision etc. To prepare and simulate these kind of solution, there are many open source frameworks like Theano, TensorFlow, CNTK, caffe, Torchnet, Deep Learning4j are available. Theano provides comparatively better hardware performance while TensorFlow provides better visualization by dataflow graphs. In this paper TensorFlow, Theano and CNTK will be compared on the basis of model Capability, Interface, Performance, Platform support, Speed, Distributed computing, Parallel execution. The best achievable goal of this work to display the best Deep Learning framework by implementing the neural network architecture for classifying images from several datasets. In above techniques some of the parameters of TensorFlow will give better performance than the others.

show abstract

“…Kovalev et al compared Theano, Torch, Caffe, Tensorflow, Keras, and DeepLearning4J in terms of speed and accuracy [4]. Shatnawi et al compared Tensorflow, Keras, Theano, and Microsoft's CNTK in terms of performance for object recognition from images [5]. Bahrampour et al compared Caffe, Neon, Tensorflow, Theano, and Torch in terms of extensibility, hardware utilization and speed for implementing convolutional and recurrent neural networks [6].…”

Section: Introductionmentioning

confidence: 99%

Makine Öğrenmesi ve Biyoenformatik Uygulamalarının Yüksek Başarımlı Hesaplama Sistemlerindeki Performans Analizi

Aydın¹

2020

Academic Platform Journal of Engineering and Science

View full text Add to dashboard Cite

Nowadays, it is becoming increasingly important to use the most efficient and most suitable computational resources for algorithmic tools that extract meaningful information from big data and make smart decisions. In this paper, a comparative analysis is provided for performance measurements of various machine learning and bioinformatics software including scikitlearn, Tensorflow, WEKA, libSVM, ThunderSVM, GMTK, PSI-BLAST, and HHblits with big data applications on different high performance computer systems and workstations. The programs are executed in a wide range of conditions such as singlecore central processing unit (CPU), multi-core CPU, and graphical processing unit (GPU) depending on the availability of implementation. The optimum number of CPU cores are obtained for selected software. It is found that the running times depend on many factors including the CPU/GPU version, available RAM, the number of CPU cores allocated, and the algorithm used. If parallel implementations are available for a given software, the best running times are typically obtained by GPU, followed by multi-core CPU, and single-core CPU. Though there is no best system that performs better than others in all applications studied, it is anticipated that the results obtained will help researchers and practitioners to select the most appropriate computational resources for their machine learning and bioinformatics projects.

show abstract

A comparative study of open source deep learning frameworks

Cited by 49 publications

References 4 publications

Performance Analysis of Deep Learning Libraries: TensorFlow and PyTorch

Performance Analysis of Deep Learning Libraries: TensorFlow and PyTorch

Review: Comparative Analysis of Different Techniques of DL-Frameworks

Makine Öğrenmesi ve Biyoenformatik Uygulamalarının Yüksek Başarımlı Hesaplama Sistemlerindeki Performans Analizi

Contact Info

Product

Resources

About