Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387561390> ?p ?o ?g. }
Showing items 1 to 73 of
73
with 100 items per page.
- W4387561390 abstract "In this paper, with rigorous derivations we formally introduce the usage of reinforcement learning to the field of inverse problems by designing an iterative algorithm, called REINFORCE-IP, for solving a general type of non-linear inverse problems. By setting specific probability densities of the action rule, we connect our approach to the conventional regularization methods of Tikhonov regularization and iterative regularization. For the numerical implementation of our approach, we parameterize the solution-searching rule with the help of neural networks and iteratively improve the parameter using a reinforcement learning algorithm -- REINFORCE. Under standard assumptions we prove the almost sure convergence of the parameter to a locally optimal value. Our work provides two typical examples (nonlinear integral equations and parameter identification problems in partial differential equations) on how reinforcement learning can be applied in solving non-linear inverse problems. Our numerical experiments show that REINFORCE-IP is an efficient algorithm that can escape from local minimums and identify multi-solutions for inverse problems with non-uniqueness." @default.
- W4387561390 created "2023-10-12" @default.
- W4387561390 creator A5007953799 @default.
- W4387561390 creator A5030624932 @default.
- W4387561390 creator A5046112287 @default.
- W4387561390 date "2023-10-10" @default.
- W4387561390 modified "2023-10-13" @default.
- W4387561390 title "Solving Inverse Problems with REINFORCE" @default.
- W4387561390 doi "https://doi.org/10.48550/arxiv.2310.06711" @default.
- W4387561390 hasPublicationYear "2023" @default.
- W4387561390 type Work @default.
- W4387561390 citedByCount "0" @default.
- W4387561390 crossrefType "posted-content" @default.
- W4387561390 hasAuthorship W4387561390A5007953799 @default.
- W4387561390 hasAuthorship W4387561390A5030624932 @default.
- W4387561390 hasAuthorship W4387561390A5046112287 @default.
- W4387561390 hasBestOaLocation W43875613901 @default.
- W4387561390 hasConcept C11413529 @default.
- W4387561390 hasConcept C121332964 @default.
- W4387561390 hasConcept C126255220 @default.
- W4387561390 hasConcept C134306372 @default.
- W4387561390 hasConcept C135252773 @default.
- W4387561390 hasConcept C152442038 @default.
- W4387561390 hasConcept C154945302 @default.
- W4387561390 hasConcept C158622935 @default.
- W4387561390 hasConcept C162324750 @default.
- W4387561390 hasConcept C207467116 @default.
- W4387561390 hasConcept C2524010 @default.
- W4387561390 hasConcept C2776135515 @default.
- W4387561390 hasConcept C2777021972 @default.
- W4387561390 hasConcept C2777303404 @default.
- W4387561390 hasConcept C28826006 @default.
- W4387561390 hasConcept C33923547 @default.
- W4387561390 hasConcept C41008148 @default.
- W4387561390 hasConcept C50522688 @default.
- W4387561390 hasConcept C50644808 @default.
- W4387561390 hasConcept C62520636 @default.
- W4387561390 hasConceptScore W4387561390C11413529 @default.
- W4387561390 hasConceptScore W4387561390C121332964 @default.
- W4387561390 hasConceptScore W4387561390C126255220 @default.
- W4387561390 hasConceptScore W4387561390C134306372 @default.
- W4387561390 hasConceptScore W4387561390C135252773 @default.
- W4387561390 hasConceptScore W4387561390C152442038 @default.
- W4387561390 hasConceptScore W4387561390C154945302 @default.
- W4387561390 hasConceptScore W4387561390C158622935 @default.
- W4387561390 hasConceptScore W4387561390C162324750 @default.
- W4387561390 hasConceptScore W4387561390C207467116 @default.
- W4387561390 hasConceptScore W4387561390C2524010 @default.
- W4387561390 hasConceptScore W4387561390C2776135515 @default.
- W4387561390 hasConceptScore W4387561390C2777021972 @default.
- W4387561390 hasConceptScore W4387561390C2777303404 @default.
- W4387561390 hasConceptScore W4387561390C28826006 @default.
- W4387561390 hasConceptScore W4387561390C33923547 @default.
- W4387561390 hasConceptScore W4387561390C41008148 @default.
- W4387561390 hasConceptScore W4387561390C50522688 @default.
- W4387561390 hasConceptScore W4387561390C50644808 @default.
- W4387561390 hasConceptScore W4387561390C62520636 @default.
- W4387561390 hasLocation W43875613901 @default.
- W4387561390 hasOpenAccess W4387561390 @default.
- W4387561390 hasPrimaryLocation W43875613901 @default.
- W4387561390 hasRelatedWork W2050033254 @default.
- W4387561390 hasRelatedWork W2127000180 @default.
- W4387561390 hasRelatedWork W2152224705 @default.
- W4387561390 hasRelatedWork W2322955667 @default.
- W4387561390 hasRelatedWork W2337734184 @default.
- W4387561390 hasRelatedWork W2373176546 @default.
- W4387561390 hasRelatedWork W2374214022 @default.
- W4387561390 hasRelatedWork W2385735574 @default.
- W4387561390 hasRelatedWork W2388364587 @default.
- W4387561390 hasRelatedWork W2561531189 @default.
- W4387561390 isParatext "false" @default.
- W4387561390 isRetracted "false" @default.
- W4387561390 workType "article" @default.