Matches in SemOpenAlex for { <https://semopenalex.org/work/W93598351> ?p ?o ?g. }
Showing items 1 to 87 of
87
with 100 items per page.
- W93598351 endingPage "487" @default.
- W93598351 startingPage "446" @default.
- W93598351 abstract "A number of important problems in theoretical computer science and machine learning can be interpreted as recovering a certain basis. These include symmetric matrix eigendecomposition, certain tensor decompositions, Independent Component Analysis (ICA), spectral clustering and Gaussian mixture learning. Each of these problems reduces to an instance of our general model, which we call a Basis Encoding Function (BEF). We show that learning a basis within this model can then be provably and efficiently achieved using a first order iteration algorithm (gradient iteration). Our algorithm goes beyond tensor methods while generalizing a number of existing algorithms---e.g., the power method for symmetric matrices, the tensor power iteration for orthogonal decomposable tensors, and cumulant-based FastICA---all within a broader function-based dynamical systems framework. Our framework also unifies the unusual phenomenon observed in these domains that they can be solved using efficient non-convex optimization. Specifically, we describe a class of BEFs such that their local maxima on the unit sphere are in one-to-one correspondence with the basis elements. This description relies on a certain hidden convexity property of these functions. We provide a complete theoretical analysis of the gradient iteration even when the BEF is perturbed. We show convergence and complexity bounds polynomial in dimension and other relevant parameters, such as perturbation size. Our perturbation results can be considered as a non-linear version of the classical Davis-Kahan theorem for perturbations of eigenvectors of symmetric matrices. In addition we show that our algorithm exhibits fast (superlinear) convergence and relate the speed of convergence to the properties of the BEF. Moreover, the gradient iteration algorithm can be easily and efficiently implemented in practice." @default.
- W93598351 created "2016-06-24" @default.
- W93598351 creator A5014532691 @default.
- W93598351 creator A5021787574 @default.
- W93598351 creator A5090342129 @default.
- W93598351 date "2016-06-06" @default.
- W93598351 modified "2023-09-27" @default.
- W93598351 title "Basis Learning as an Algorithmic Primitive" @default.
- W93598351 hasPublicationYear "2016" @default.
- W93598351 type Work @default.
- W93598351 sameAs 93598351 @default.
- W93598351 citedByCount "0" @default.
- W93598351 crossrefType "proceedings-article" @default.
- W93598351 hasAuthorship W93598351A5014532691 @default.
- W93598351 hasAuthorship W93598351A5021787574 @default.
- W93598351 hasAuthorship W93598351A5090342129 @default.
- W93598351 hasConcept C105795698 @default.
- W93598351 hasConcept C11413529 @default.
- W93598351 hasConcept C121332964 @default.
- W93598351 hasConcept C12426560 @default.
- W93598351 hasConcept C126255220 @default.
- W93598351 hasConcept C134306372 @default.
- W93598351 hasConcept C154945302 @default.
- W93598351 hasConcept C155281189 @default.
- W93598351 hasConcept C158693339 @default.
- W93598351 hasConcept C159694833 @default.
- W93598351 hasConcept C162443888 @default.
- W93598351 hasConcept C169756996 @default.
- W93598351 hasConcept C202444582 @default.
- W93598351 hasConcept C2524010 @default.
- W93598351 hasConcept C28826006 @default.
- W93598351 hasConcept C33923547 @default.
- W93598351 hasConcept C41008148 @default.
- W93598351 hasConcept C51432778 @default.
- W93598351 hasConcept C5917680 @default.
- W93598351 hasConcept C62520636 @default.
- W93598351 hasConcept C73555534 @default.
- W93598351 hasConceptScore W93598351C105795698 @default.
- W93598351 hasConceptScore W93598351C11413529 @default.
- W93598351 hasConceptScore W93598351C121332964 @default.
- W93598351 hasConceptScore W93598351C12426560 @default.
- W93598351 hasConceptScore W93598351C126255220 @default.
- W93598351 hasConceptScore W93598351C134306372 @default.
- W93598351 hasConceptScore W93598351C154945302 @default.
- W93598351 hasConceptScore W93598351C155281189 @default.
- W93598351 hasConceptScore W93598351C158693339 @default.
- W93598351 hasConceptScore W93598351C159694833 @default.
- W93598351 hasConceptScore W93598351C162443888 @default.
- W93598351 hasConceptScore W93598351C169756996 @default.
- W93598351 hasConceptScore W93598351C202444582 @default.
- W93598351 hasConceptScore W93598351C2524010 @default.
- W93598351 hasConceptScore W93598351C28826006 @default.
- W93598351 hasConceptScore W93598351C33923547 @default.
- W93598351 hasConceptScore W93598351C41008148 @default.
- W93598351 hasConceptScore W93598351C51432778 @default.
- W93598351 hasConceptScore W93598351C5917680 @default.
- W93598351 hasConceptScore W93598351C62520636 @default.
- W93598351 hasConceptScore W93598351C73555534 @default.
- W93598351 hasLocation W935983511 @default.
- W93598351 hasOpenAccess W93598351 @default.
- W93598351 hasPrimaryLocation W935983511 @default.
- W93598351 hasRelatedWork W1563366202 @default.
- W93598351 hasRelatedWork W1578157337 @default.
- W93598351 hasRelatedWork W1650849357 @default.
- W93598351 hasRelatedWork W2090859773 @default.
- W93598351 hasRelatedWork W2092332806 @default.
- W93598351 hasRelatedWork W2390688333 @default.
- W93598351 hasRelatedWork W2401335853 @default.
- W93598351 hasRelatedWork W2544583424 @default.
- W93598351 hasRelatedWork W2737605135 @default.
- W93598351 hasRelatedWork W2738237501 @default.
- W93598351 hasRelatedWork W2739509141 @default.
- W93598351 hasRelatedWork W2756330084 @default.
- W93598351 hasRelatedWork W2757782085 @default.
- W93598351 hasRelatedWork W2950168472 @default.
- W93598351 hasRelatedWork W2970050564 @default.
- W93598351 hasRelatedWork W2996740905 @default.
- W93598351 hasRelatedWork W3006821019 @default.
- W93598351 hasRelatedWork W3034046696 @default.
- W93598351 hasRelatedWork W3123821778 @default.
- W93598351 hasRelatedWork W52203363 @default.
- W93598351 isParatext "false" @default.
- W93598351 isRetracted "false" @default.
- W93598351 magId "93598351" @default.
- W93598351 workType "article" @default.