Open main menu

Wikipedia β

Comparison of deep learning software

The following table compares some of the most popular software frameworks, libraries and computer programs for deep learning.


Deep learning software by nameEdit

Software Creator Software license[a] Open source Platform Written in Interface OpenMP support OpenCL support CUDA support Automatic differentiation[1] Has pretrained models Recurrent nets Convolutional nets RBM/DBNs Parallel execution (multi node) Kevin Lok MIT Yes Linux, macOS, Windows Python Python Yes Yes Yes Yes
BigDL Jason Dai Apache 2.0 Yes Apache Spark Scala Scala, Python No Yes Yes Yes
Caffe Berkeley Vision and Learning Center BSD license Yes Linux, macOS, Windows[2] C++ Python, MATLAB, C++ Yes Under development[3] Yes Yes Yes[4] Yes Yes No ?
Deeplearning4j Skymind engineering team; Deeplearning4j community; originally Adam Gibson Apache 2.0 Yes Linux, macOS, Windows, Android (Cross-platform) C++, Java Java, Scala, Clojure, Python (Keras), Kotlin Yes On roadmap[5] Yes[6][7] Computational Graph Yes[8] Yes Yes Yes Yes[9]
Chainer Preferred Networks MIT license Yes Linux, macOS, Windows Python No Yes Yes Yes Yes Yes Yes
Darknet Joseph Redmon Public Domain Yes Cross-Platform C C, Python Yes Yes Yes Yes
Dlib Davis King Boost Software License Yes Cross-Platform C++ C++ Yes No Yes Yes Yes No Yes Yes Yes
DataMelt (DMelt) S.Chekanov Freemium Yes Cross-Platform Java Java No No No No No No No No No
DyNet Carnegie Mellon University Apache 2.0 Yes Linux, macOS, Windows C++, Python Yes Yes Yes Yes
Intel Data Analytics Acceleration Library Intel Apache License 2.0 Yes Linux, macOS, Windows on Intel CPU[10] C++, Python, Java C++, Python, Java[10] Yes No No Yes No Yes Yes
Intel Math Kernel Library Intel Proprietary No Linux, macOS, Windows on Intel CPU[11] C[12] Yes[13] No No Yes No Yes[14] Yes[14] No
Keras François Chollet MIT license Yes Linux, macOS, Windows Python Python, R Only if using Theano as backend Under development for the Theano backend (and on roadmap for the TensorFlow backend) Yes Yes Yes[15] Yes Yes Yes Yes[16]
MATLAB + Neural Network Toolbox MathWorks Proprietary No Linux, macOS, Windows C, C++, Java, MATLAB MATLAB No No Train with Parallel Computing Toolbox and generate CUDA code with GPU Coder[17] No Yes[18][19] Yes[18] Yes[18] No With Parallel Computing Toolbox[20]
Microsoft Cognitive Toolkit Microsoft Research MIT license[21] Yes Windows, Linux[22] (macOS via Docker on roadmap) C++ Python (Keras), C++, Command line,[23] BrainScript[24] (.NET on roadmap[25]) Yes[26] No Yes Yes Yes[27] Yes[28] Yes[28] No[29] Yes[30]
Apache MXNet Apache Software Foundation Apache 2.0 Yes Linux, macOS, Windows,[31][32] AWS, Android,[33] iOS, JavaScript[34] Small C++ core library C++, Python, Julia, Matlab, JavaScript, Go, R, Scala, Perl Yes On roadmap[35] Yes Yes[36] Yes[37] Yes Yes Yes Yes[38]
Neural Designer Artelnics Proprietary No Linux, macOS, Windows C++ Graphical user interface Yes No No ? ? No No No ?
OpenNN Artelnics GNU LGPL Yes Cross-platform C++ C++ Yes No Yes ? ? No No No ?
PyTorch Adam Paszke, Sam Gross, Soumith Chintala, Gregory Chanan BSD license Yes Linux, macOS, Windows Python, C, CUDA Python Yes Yes Yes Yes Yes Yes Yes
Apache SINGA Apache Incubator Apache 2.0 Yes Linux, macOS, Windows C++ Python, C++, Java No No Yes ? Yes Yes Yes Yes Yes
TensorFlow Google Brain team Apache 2.0 Yes Linux, macOS, Windows,[39] Android C++, Python, CUDA Python (Keras), C/C++, Java, Go, R[40], Julia No On roadmap[41] but already with SYCL[42] support Yes Yes[43] Yes[44] Yes Yes Yes Yes
Theano Université de Montréal BSD license Yes Cross-platform Python Python (Keras) Yes Under development[45] Yes Yes[46][47] Through Lasagne's model zoo[48] Yes Yes Yes Yes[49]
Torch Ronan Collobert, Koray Kavukcuoglu, Clement Farabet BSD license Yes Linux, macOS, Windows,[50] Android,[51] iOS C, Lua Lua, LuaJIT,[52] C, utility library for C++/OpenCL[53] Yes Third party implementations[54][55] Yes[56][57] Through Twitter's Autograd[58] Yes[59] Yes Yes Yes Yes[60]
Wolfram Mathematica Wolfram Research Proprietary No Windows, macOS, Linux, Cloud computing C++ Wolfram Language No No Yes Yes Yes[61] Yes Yes Yes Yes
  1. ^ Licenses here are a summary, and are not taken to be complete statements of the licenses. Some libraries may use other libraries internally under different licenses

Related softwareEdit

See alsoEdit


  1. ^ Atilim Gunes Baydin; Barak A. Pearlmutter; Alexey Andreyevich Radul; Jeffrey Mark Siskind (20 February 2015). "Automatic differentiation in machine learning: a survey". arXiv:1502.05767  [cs.LG]. 
  2. ^ "Microsoft/caffe". GitHub. 
  3. ^ "OpenCL Caffe". 
  4. ^ "Caffe Model Zoo". 
  5. ^ "Support for Open CL · Issue #27 · deeplearning4j/nd4j". GitHub. 
  6. ^ "N-Dimensional Scientific Computing for Java". 
  7. ^ "Comparing Top Deep Learning Frameworks". Deeplearning4j. 
  8. ^ Chris Nicholson; Adam Gibson. "Deeplearning4j Models". 
  9. ^ Deeplearning4j. "Deeplearning4j on Spark". Deeplearning4j. 
  10. ^ a b Intel® Data Analytics Acceleration Library (Intel® DAAL) | Intel® Software
  11. ^ Intel® Math Kernel Library (Intel® MKL) | Intel® Software
  12. ^ Deep Neural Network Functions
  13. ^ Using Intel® MKL with Threaded Applications | Intel® Software
  14. ^ a b Intel® Xeon Phi™ Delivers Competitive Performance For Deep Learning—And Getting Better Fast | Intel® Software
  15. ^
  16. ^ Does Keras support using multiple GPUs? · Issue #2436 · fchollet/keras
  17. ^ "GPU Coder - MATLAB & Simulink". MathWorks. Retrieved 13 November 2017. 
  18. ^ a b c "Neural Network Toolbox - MATLAB". MathWorks. Retrieved 13 November 2017. 
  19. ^ "Deep Learning Models - MATLAB & Simulink". MathWorks. Retrieved 13 November 2017. 
  20. ^ "Parallel Computing Toolbox - MATLAB". MathWorks. Retrieved 13 November 2017. 
  21. ^ "CNTK/ at master · Microsoft/CNTK · GitHub". GitHub. 
  22. ^ "Setup CNTK on your machine". GitHub. 
  23. ^ "CNTK usage overview". GitHub. 
  24. ^ "BrainScript Network Builder". GitHub. 
  25. ^ ".NET Support · Issue #960 · Microsoft/CNTK". GitHub. 
  26. ^ "How to train a model using multiple machines? · Issue #59 · Microsoft/CNTK". GitHub. 
  27. ^
  28. ^ a b "CNTK - Computational Network Toolkit". Microsoft Corporation. 
  29. ^ url=
  30. ^ "Multiple GPUs and machines". Microsoft Corporation. 
  31. ^ "Releases · dmlc/mxnet". Github. 
  32. ^ "Installation Guide — mxnet documentation". Readthdocs. 
  33. ^ "MXNet Smart Device". ReadTheDocs. 
  34. ^ "MXNet.js". Github. 
  35. ^ "Support for other Device Types, OpenCL AMD GPU · Issue #621 · dmlc/mxnet". GitHub. 
  36. ^
  37. ^ "Model Gallery". GitHub. 
  38. ^ "Run MXNet on Multiple CPU/GPUs with Data Parallel". GitHub. 
  39. ^
  40. ^ interface), JJ Allaire (R; RStudio; Eddelbuettel, Dirk; Golding, Nick; Tang, Yuan; Tutorials), Google Inc (Examples and (2017-05-26), tensorflow: R Interface to TensorFlow, retrieved 2017-06-14 
  41. ^ "tensorflow/ at master · tensorflow/tensorflow · GitHub". GitHub. January 23, 2017. Retrieved May 21, 2017. 
  42. ^ "OpenCL support · Issue #22 · tensorflow/tensorflow". GitHub. 
  43. ^
  44. ^
  45. ^ "Using the GPU — Theano 0.8.2 documentation". 
  46. ^
  47. ^
  48. ^ "Recipes/modelzoo at master · Lasagne/Recipes · GitHub". GitHub. 
  49. ^ Using multiple GPUs — Theano 0.8.2 documentation
  50. ^
  51. ^ "GitHub - soumith/torch-android: Torch-7 for Android". GitHub. 
  52. ^ "Torch7: A Matlab-like Environment for Machine Learning" (PDF). 
  53. ^ "GitHub - jonathantompson/jtorch: An OpenCL Torch Utility Library". GitHub. 
  54. ^ "Cheatsheet". GitHub. 
  55. ^ "cltorch". GitHub. 
  56. ^ "Torch CUDA backend". GitHub. 
  57. ^ "Torch CUDA backend for nn". GitHub. 
  58. ^
  59. ^ "ModelZoo". GitHub. 
  60. ^
  61. ^