FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3167630251> ?p ?o ?g. }

• W3167630251 abstract "Automatic differentiation (autodiff) has revolutionized machine learning. It allows expressing complex computations by composing elementary ones in creative ways and removes the burden of computing their derivatives by hand. More recently, differentiation of optimization problem solutions has attracted widespread attention with applications such as optimization as a layer, and in bi-level problems such as hyper-parameter optimization and meta-learning. However, the formulas for these derivatives often involve case-by-case tedious mathematical derivations. In this paper, we propose a unified, efficient and modular approach for implicit differentiation of optimization problems. In our approach, the user defines (in Python in the case of our implementation) a function $F$ capturing the optimality conditions of the problem to be differentiated. Once this is done, we leverage autodiff of $F$ and implicit differentiation to automatically differentiate the optimization problem. Our approach thus combines the benefits of implicit differentiation and autodiff. It is efficient as it can be added on top of any state-of-the-art solver and modular as the optimality condition specification is decoupled from the implicit differentiation mechanism. We show that seemingly simple principles allow to recover many recently proposed implicit differentiation methods and create new ones easily. We demonstrate the ease of formulating and solving bi-level optimization problems using our framework. We also showcase an application to the sensitivity analysis of molecular dynamics." @default.
• W3167630251 created "2021-06-22" @default.
• W3167630251 creator A5017263427 @default.
• W3167630251 creator A5024896990 @default.
• W3167630251 creator A5049123454 @default.
• W3167630251 creator A5051850054 @default.
• W3167630251 creator A5064770739 @default.
• W3167630251 creator A5066572762 @default.
• W3167630251 creator A5069963995 @default.
• W3167630251 creator A5070161680 @default.
• W3167630251 date "2021-05-31" @default.
• W3167630251 modified "2023-09-27" @default.
• W3167630251 title "Efficient and Modular Implicit Differentiation." @default.
• W3167630251 cites W1522301498 @default.
• W3167630251 cites W1542343170 @default.
• W3167630251 cites W1585773866 @default.
• W3167630251 cites W1601741115 @default.
• W3167630251 cites W163279865 @default.
• W3167630251 cites W1953936588 @default.
• W3167630251 cites W1978119584 @default.
• W3167630251 cites W1978259121 @default.
• W3167630251 cites W2044028691 @default.
• W3167630251 cites W2052387674 @default.
• W3167630251 cites W2056385679 @default.
• W3167630251 cites W2058760524 @default.
• W3167630251 cites W2063318134 @default.
• W3167630251 cites W2063978378 @default.
• W3167630251 cites W2093781566 @default.
• W3167630251 cites W2101234009 @default.
• W3167630251 cites W2121729458 @default.
• W3167630251 cites W2140153041 @default.
• W3167630251 cites W2153838241 @default.
• W3167630251 cites W2155216327 @default.
• W3167630251 cites W2157791002 @default.
• W3167630251 cites W2158001550 @default.
• W3167630251 cites W2158131535 @default.
• W3167630251 cites W2163112044 @default.
• W3167630251 cites W2166107799 @default.
• W3167630251 cites W2187061624 @default.
• W3167630251 cites W2288174618 @default.
• W3167630251 cites W2409290336 @default.
• W3167630251 cites W2505728881 @default.
• W3167630251 cites W2751681791 @default.
• W3167630251 cites W2842089854 @default.
• W3167630251 cites W2884724578 @default.
• W3167630251 cites W2899971035 @default.
• W3167630251 cites W2913535645 @default.
• W3167630251 cites W2939853902 @default.
• W3167630251 cites W2949847915 @default.
• W3167630251 cites W2962736834 @default.
• W3167630251 cites W2962826047 @default.
• W3167630251 cites W2963123301 @default.
• W3167630251 cites W2963130368 @default.
• W3167630251 cites W2963306862 @default.
• W3167630251 cites W2963502387 @default.
• W3167630251 cites W2963804140 @default.
• W3167630251 cites W2963842222 @default.
• W3167630251 cites W2963970238 @default.
• W3167630251 cites W2964095938 @default.
• W3167630251 cites W2968953580 @default.
• W3167630251 cites W2970971581 @default.
• W3167630251 cites W2971296616 @default.
• W3167630251 cites W2971842688 @default.
• W3167630251 cites W2972419151 @default.
• W3167630251 cites W2972954809 @default.
• W3167630251 cites W2981603589 @default.
• W3167630251 cites W2995268367 @default.
• W3167630251 cites W3022909171 @default.
• W3167630251 cites W3034232789 @default.
• W3167630251 cites W3034508441 @default.
• W3167630251 cites W3035212372 @default.
• W3167630251 cites W3035462522 @default.
• W3167630251 cites W3037120332 @default.
• W3167630251 cites W3092802849 @default.
• W3167630251 cites W3100611487 @default.
• W3167630251 cites W3101933250 @default.
• W3167630251 cites W3106216259 @default.
• W3167630251 cites W3106375145 @default.
• W3167630251 cites W3130740406 @default.
• W3167630251 cites W3157908369 @default.
• W3167630251 cites W3162555546 @default.
• W3167630251 cites W2290452516 @default.
• W3167630251 hasPublicationYear "2021" @default.
• W3167630251 type Work @default.
• W3167630251 sameAs 3167630251 @default.
• W3167630251 citedByCount "5" @default.
• W3167630251 countsByYear W31676302512021 @default.
• W3167630251 countsByYear W31676302512022 @default.
• W3167630251 crossrefType "posted-content" @default.
• W3167630251 hasAuthorship W3167630251A5017263427 @default.
• W3167630251 hasAuthorship W3167630251A5024896990 @default.
• W3167630251 hasAuthorship W3167630251A5049123454 @default.
• W3167630251 hasAuthorship W3167630251A5051850054 @default.
• W3167630251 hasAuthorship W3167630251A5064770739 @default.
• W3167630251 hasAuthorship W3167630251A5066572762 @default.
• W3167630251 hasAuthorship W3167630251A5069963995 @default.
• W3167630251 hasAuthorship W3167630251A5070161680 @default.
• W3167630251 hasConcept C101468663 @default.
• W3167630251 hasConcept C111472728 @default.
• W3167630251 hasConcept C11413529 @default.

• next