%0 Journal Article %1 elliott2018simple %A Elliott, Conal %D 2018 %K autograd deep-learning optimization %T The simple essence of automatic differentiation %U http://arxiv.org/abs/1804.00746 %X Automatic differentiation (AD) in reverse mode (RAD) is a central component of deep learning and other uses of large-scale optimization. Commonly used RAD algorithms such as backpropagation, however, are complex and stateful, hindering deep understanding, improvement, and parallel execution. This paper develops a simple, generalized AD algorithm calculated from a simple, natural specification. The general algorithm is then specialized by varying the representation of derivatives. In particular, applying well-known constructions to a naive representation yields two RAD algorithms that are far simpler than previously known. In contrast to commonly used RAD implementations, the algorithms defined here involve no graphs, tapes, variables, partial derivatives, or mutation. They are inherently parallel-friendly, correct by construction, and usable directly from an existing programming language with no need for new data types or programming style, thanks to use of an AD-agnostic compiler plugin.

@article{elliott2018simple, abstract = {Automatic differentiation (AD) in reverse mode (RAD) is a central component of deep learning and other uses of large-scale optimization. Commonly used RAD algorithms such as backpropagation, however, are complex and stateful, hindering deep understanding, improvement, and parallel execution. This paper develops a simple, generalized AD algorithm calculated from a simple, natural specification. The general algorithm is then specialized by varying the representation of derivatives. In particular, applying well-known constructions to a naive representation yields two RAD algorithms that are far simpler than previously known. In contrast to commonly used RAD implementations, the algorithms defined here involve no graphs, tapes, variables, partial derivatives, or mutation. They are inherently parallel-friendly, correct by construction, and usable directly from an existing programming language with no need for new data types or programming style, thanks to use of an AD-agnostic compiler plugin.}, added-at = {2019-07-18T16:54:13.000+0200}, author = {Elliott, Conal}, biburl = {https://www.bibsonomy.org/bibtex/2603b99f11a41903b320a035dc320e3f9/kirk86}, description = {[1804.00746] The simple essence of automatic differentiation}, interhash = {2aa4b22867eca50ab696b8cdcf744eaf}, intrahash = {603b99f11a41903b320a035dc320e3f9}, keywords = {autograd deep-learning optimization}, note = {cite arxiv:1804.00746Comment: 37 pages with proof appendices and 15 figures. Extended version of a paper appearing at ICFP 2018. More info at http://conal.net/papers/essence-of-ad/}, timestamp = {2019-07-18T16:54:13.000+0200}, title = {The simple essence of automatic differentiation}, url = {http://arxiv.org/abs/1804.00746}, year = 2018 }