Flux (machine-learning framework)

Flux is an open-source machine-learning software library and ecosystem written in Julia.[1][3] Its current stable release is v0.10.3.[4] It has a layer-stacking-based interface for simpler models, and has a strong support on interoperability with other Julia packages instead of a monolithic design.[5] For example, GPU support is implemented transparently by CuArrays.jl[6] This is in contrast to some other machine learning frameworks which are implemented in other languages with Julia bindings, such as TensorFlow.jl, and thus are more limited by the functionality present in the underlying implementation, which is often in C or C++.[7]

Flux
Original author(s)Michael J Innes.[1]
Stable release
v0.10.3
Repositorygithub.com/FluxML/Flux.jl
Written inJulia
TypeMachine learning library
LicenseMIT[2]
Websitehttps://fluxml.ai

Flux's focus on interoperability has enabled, for example, support for Neural Differential Equations, by fusing Flux.jl and DifferentialEquations.jl into DiffEqFlux.jl.[8][9]

Flux supports recurrent and convolutional networks. It is also capable of differentiable programming[10][11][12] through its source-to-source automatic differentiation package, Zygote.jl.[13]

Julia is a popular language in machine-learning[14] and Flux.jl is its most highly regarded machine-learning repository.[14] A demonstration[15] compiling Julia code to run in Google's Tensor processing unit received praise from Google Brain AI lead Jeff Dean.[16]

Flux has been used as a framework to build neural networks that work with homomorphic encrypted data without ever decrypting it.[17][18] This kind of application is envisioned to be central for privacy to future API using machine-learning models.[19]

Flux.jl is an intermediate representation for running high level programs on CUDA hardware.[20][21] It was the predecessor to CUDAnative.jl which is also a GPU programming language.[22]

See also

References
  1. Innes, Michael (2018-05-03). "Flux: Elegant machine learning with Julia". Journal of Open Source Software. 3 (25): 602. doi:10.21105/joss.00602.
  2. "github.com/FluxML/Flux.jl/blob/master/LICENSE.md".
  3. Innes, Mike; Bradbury, James; Fischer, Keno; Gandhi, Dhairya; Mariya Joy, Neethu; Karmali, Tejan; Kelley, Matt; Pal, Avik; Concetto Rudilosso, Marco; Saba, Elliot; Shah, Viral; Yuret, Deniz. "Building a Language and Compiler for Machine Learning". julialang.org. Retrieved 2019-06-02.
  4. FluxML/Flux.jl v0.10.3, Flux, 2020-03-04, retrieved 2020-03-27
  5. "Machine Learning and Artificial Intelligence". juliacomputing.com. Retrieved 2019-06-02.
  6. Gandhi, Dhairya (2018-11-15). "Julia at NeurIPS and the Future of Machine Learning Tools". juliacomputing.com. Retrieved 2019-06-02.
  7. Malmaud, Jonathan; White, Lyndon (2018-11-01). "TensorFlow.jl: An Idiomatic Julia Front End for TensorFlow". Journal of Open Source Software. 3 (31): 1002. doi:10.21105/joss.01002.
  8. Rackauckas, Chris; Innes, Mike; Ma, Yingbo; Bettencourt, Jesse; White, Lyndon; Dixit, Vaibhav (2019-02-06). "DiffEqFlux.jl - A Julia Library for Neural Differential Equations". arXiv:1902.02376 [cs.LG].
  9. Schlothauer, Sarah (2019-01-25). "Machine learning meets math: Solve differential equations with new Julia library". JAXenter. Retrieved 2019-10-21.
  10. "Flux – Reinforcement Learning vs. Differentiable Programming". fluxml.ai. Retrieved 2019-06-02.
  11. "Flux – What Is Differentiable Programming?". fluxml.ai. Retrieved 2019-06-02.
  12. Heath, Nick (December 6, 2018). "Julia vs Python: Which programming language will rule machine learning in 2019?". TechRepublic. Retrieved 2019-06-03.
  13. Innes, Michael (2018-10-18). "Don't Unroll Adjoint: Differentiating SSA-Form Programs". arXiv:1810.07951 [cs.PL].
  14. Heath, Nick (January 25, 2019). "GitHub: The top 10 programming languages for machine learning". TechRepublic. Retrieved 2019-06-03.
  15. Saba, Elliot; Fischer, Keno (2018-10-23). "Automatic Full Compilation of Julia Programs and ML Models to Cloud TPUs". arXiv:1810.09868 [cs.PL].
  16. Dean, Jeff [@JeffDean] (2018-10-23). "Julia + TPUs = fast and easily expressible ML computations" (Tweet). Retrieved 2019-06-02 – via Twitter.
  17. Patrawala, Fatema (2019-11-28). "Julia Computing research team runs machine learning model on encrypted data without decrypting it". Packt Hub. Retrieved 2019-12-11.
  18. "Machine Learning on Encrypted Data Without Decrypting It". juliacomputing.com. 2019-11-22. Retrieved 2019-12-11.
  19. Yadav, Rohit (2019-12-02). "Julia Computing Uses Homomorphic Encryption For ML. Is It The Way Forward?". Analytics India Magazine. Retrieved 2019-12-11.
  20. Roesch, Jared and Lyubomirsky, Steven and Kirisame, Marisa and Pollock, Josh and Weber, Logan and Jiang, Ziheng and Chen, Tianqi and Moreau, Thierry and Tatlock, Zachary (2019). "Relay: A High-Level IR for Deep Learning". arXiv:1904.08368.CS1 maint: multiple names: authors list (link)
  21. Tim Besard and Christophe Foket and Bjorn De Sutter (2019). "Effective Extensible Programming: Unleashing Julia on GPUs". IEEE Transactions on Parallel and Distributed Systems. Institute of Electrical and Electronics Engineers (IEEE). 30 (4): 827–841. arXiv:1712.03112. doi:10.1109/tpds.2018.2872064.
  22. Besard, Tim (2018). Abstractions for Programming Graphics Processors in High-Level Programming Languages (PhD). Ghent University.
This article is issued from Wikipedia. The text is licensed under Creative Commons - Attribution - Sharealike. Additional terms may apply for the media files.