Differentiable programming: Difference between revisions

Revision as of 07:56, 13 February 2019

Revision as of 08:07, 13 February 2019

Line 1:

'''Differentiable programming''' is a [[programming paradigm]] in which the entire programprograms can be [[Differentiation (mathematics)|differentiated]] throughout, usually via [[automatic differentiation]].<ref>{{Citation|last=Wang|first=Fei|title=Backpropagation with Callbacks: Foundations for Efficient and Expressive Differentiable Programming|date=2018|url=http://papers.nips.cc/paper/8221-backpropagation-with-callbacks-foundations-for-efficient-and-expressive-differentiable-programming.pdf|work=Advances in Neural Information Processing Systems 31|pages=10201–10212|editor-last=Bengio|editor-first=S.|publisher=Curran Associates, Inc.|access-date=2019-02-13|last2=Decker|first2=James|last3=Wu|first3=Xilun|last4=Essertel|first4=Gregory|last5=Rompf|first5=Tiark|editor2-last=Wallach|editor2-first=H.|editor3-last=Larochelle|editor3-first=H.|editor4-last=Grauman|editor4-first=K.}}</ref><ref name="innes">{{Cite journal|last=Innes|first=Mike|date=2018|title=On Machine Learning and Programming Languages|url=http://www.sysml.cc/doc/37.pdf|journal=SysML Conference 2018|volume=|pages=|via=}}</ref> This allows for [[Gradient method|gradient based optimization]] of parameters in the program, often via [[gradient descent]]. Differentiable programming has found use in areas such as combining [[deep learning]] with [[physics engines]] in [[robotics]], differentiable [[Ray tracing (graphics)|ray tracing]], and [[image processing]].<ref>{{Cite journal|last=Degrave|first=Jonas|last2=Hermans|first2=Michiel|last3=Dambre|first3=Joni|last4=wyffels|first4=Francis|date=2016-11-05|title=A Differentiable Physics Engine for Deep Learning in Robotics|url=http://arxiv.org/abs/1611.01652|journal=arXiv:1611.01652 [cs]}}</ref><ref>{{Cite web|url=https://people.csail.mit.edu/tzumao/diffrt/|title=Differentiable Monte Carlo Ray Tracing through Edge Sampling|website=people.csail.mit.edu|access-date=2019-02-13}}</ref><ref>{{Cite web|url=https://people.csail.mit.edu/tzumao/gradient_halide/|title=Differentiable Programming for Image Processing and Deep Learning in Halide|website=people.csail.mit.edu|access-date=2019-02-13}}</ref>

Most differentiable programming frameworks work by constructing a graph containing the control flow and [[data structures]] in the program.<ref name="flux">{{Cite journal|last=Innes|first=Michael|last2=Saba|first2=Elliot|last3=Fischer|first3=Keno|last4=Gandhi|first4=Dhairya|last5=Rudilosso|first5=Marco Concetto|last6=Joy|first6=Neethu Mariya|last7=Karmali|first7=Tejan|last8=Pal|first8=Avik|last9=Shah|first9=Viral|date=2018-10-31|title=Fashionable Modelling with Flux|url=http://arxiv.org/abs/1811.01457|journal=arXiv:1811.01457 [cs]}}</ref> Earlier attempts generally featured a tradeoff between a "dynamic" [[Interpreted language|interpreted]] graph — chosen by frameworks such as [[PyTorch]] and [[AutoGrad (NumPy)|AutoGrad]] — which leads to interpreter overhead and poorer scalability, and a "static" [[compiled]] graph — chosen by frameworks such as [[TensorFlow]] — which limits interactivity and the types of programs that can be created easily, as well as making it harder for users to reason effectively about their programs. These earlier attempts are also generally only able to differentiate code written in a suitable manner for the framework, limiting their interoperability with other programs.<ref name="flux" /> A more recent framework in the [[Julia (programming language)|Julia]] programming language - called Zygote - resolves these problems by treating the language's syntax as the graph; the code's [[intermediate representation]] can then be differentiated directly and compiled.<ref name="flux" /><ref>{{Cite journal|last=Innes|first=Michael|date=2018-10-18|title=Don't Unroll Adjoint: Differentiating SSA-Form Programs|url=http://arxiv.org/abs/1810.07951|journal=arXiv:1810.07951 [cs]}}</ref>