Symmetric rank-one (original) (raw)
The Symmetric Rank 1 (SR1) method is a quasi-Newton method to update the second derivative (Hessian)based on the derivatives (gradients) calculated at two points. It is a generalization to the secant method for a multidimensional problem.This update maintains the symmetry of the matrix but does not guarantee that the update be positive definite. A twice continuously differentiable function has a gradient and Hessian matrix : The function has an expansion as a Taylor series at , which can be truncated ; its gradient has a Taylor-series approximation also , , where . . ,
| Property | Value |
|---|---|
| dbo:abstract | The Symmetric Rank 1 (SR1) method is a quasi-Newton method to update the second derivative (Hessian)based on the derivatives (gradients) calculated at two points. It is a generalization to the secant method for a multidimensional problem.This update maintains the symmetry of the matrix but does not guarantee that the update be positive definite. The sequence of Hessian approximations generated by the SR1 method converges to the true Hessian under mild conditions, in theory; in practice, the approximate Hessians generated by the SR1 method show faster progress towards the true Hessian than do popular alternatives (BFGS or DFP), in preliminary numerical experiments. The SR1 method has computational advantages for sparse or problems. A twice continuously differentiable function has a gradient and Hessian matrix : The function has an expansion as a Taylor series at , which can be truncated ; its gradient has a Taylor-series approximation also , which is used to update . The above secant-equation need not have a unique solution .The SR1 formula computes (via an update of rank 1) the symmetric solution that is closest to the current approximate-value : , where . The corresponding update to the approximate inverse-Hessian is . One might wonder why positive-definiteness is not preserved — after all, a rank-1 update of the form is positive-definite if is. The explanation is that the update might be of the form instead because the denominator can be negative, and in that case there are no guarantees about positive-definiteness. The SR1 formula has been rediscovered a number of times. A drawback is that the denominator can vanish. Some authors have suggested that the update be applied only if , where is a small number, e.g. . (en) |
| dbo:wikiPageID | 11227519 (xsd:integer) |
| dbo:wikiPageLength | 4324 (xsd:nonNegativeInteger) |
| dbo:wikiPageRevisionID | 1102978358 (xsd:integer) |
| dbo:wikiPageWikiLink | dbr:Quasi-Newton_method dbr:Gradient dbr:Davidon-Fletcher-Powell_formula dbr:Broyden's_method dbc:Optimization_algorithms_and_methods dbr:Rank_(linear_algebra) dbr:Taylor_series dbr:Hessian_matrix dbr:Positive_definite_matrix dbr:Newton's_method_in_optimization dbr:Secant_method dbr:BFGS dbr:BFGS_method dbr:L-BFGS dbr:Sparsity dbr:Partial_separability |
| dbp:wikiPageUsesTemplate | dbt:Explain dbt:Reflist dbt:Optimization_algorithms |
| dct:subject | dbc:Optimization_algorithms_and_methods |
| rdf:type | yago:WikicatOptimizationAlgorithmsAndMethods yago:Abstraction100002137 yago:Act100030358 yago:Activity100407535 yago:Algorithm105847438 yago:Event100029378 yago:Procedure101023820 yago:PsychologicalFeature100023100 yago:YagoPermanentlyLocatedEntity yago:Rule105846932 |
| rdfs:comment | The Symmetric Rank 1 (SR1) method is a quasi-Newton method to update the second derivative (Hessian)based on the derivatives (gradients) calculated at two points. It is a generalization to the secant method for a multidimensional problem.This update maintains the symmetry of the matrix but does not guarantee that the update be positive definite. A twice continuously differentiable function has a gradient and Hessian matrix : The function has an expansion as a Taylor series at , which can be truncated ; its gradient has a Taylor-series approximation also , , where . . , (en) |
| rdfs:label | Symmetric rank-one (en) |
| owl:sameAs | freebase:Symmetric rank-one yago-res:Symmetric rank-one wikidata:Symmetric rank-one https://global.dbpedia.org/id/4vZcv |
| prov:wasDerivedFrom | wikipedia-en:Symmetric_rank-one?oldid=1102978358&ns=0 |
| foaf:isPrimaryTopicOf | wikipedia-en:Symmetric_rank-one |
| is dbo:wikiPageRedirects of | dbr:SR1_formula |
| is dbo:wikiPageWikiLink of | dbr:SR1_formula dbr:Davidon–Fletcher–Powell_formula dbr:List_of_numerical_analysis_topics dbr:Broyden–Fletcher–Goldfarb–Shanno_algorithm dbr:SR1_(disambiguation) |
| is foaf:primaryTopic of | wikipedia-en:Symmetric_rank-one |