2012
DOI: 10.13001/1081-3810.1551
|View full text |Cite
|
Sign up to set email alerts
|

Interest zone matrix approximation

Abstract: We present an algorithm for low rank approximation of matrices where only some of the entries in the matrix are taken into consideration. This algorithm appears in recent literature under different names, where it is described as an EM based algorithm that maximizes the likelihood for the missing entries without any relation for the mean square error minimization. When the algorithm is minimized from a mean-square-error point of view, we prove that the error produced by the algorithm is monotonically decreasin… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
2

Citation Types

0
15
1
3

Year Published

2015
2015
2023
2023

Publication Types

Select...
3
2
1
1

Relationship

1
6

Authors

Journals

citations
Cited by 13 publications
(19 citation statements)
references
References 17 publications
0
15
1
3
Order By: Relevance
“…The algorithms return a local minimum (P M ) or a completion-valid point (P C ) by the application of a stochastic gradient descent (or other optimizer), but does not guarantee to return the closest point, since it depends on the structure of the neural network, which is typically high-dimensional non-convnex manifold. This is different than the case in [22].…”
Section: Theoretical Analysiscontrasting
confidence: 72%
See 1 more Smart Citation
“…The algorithms return a local minimum (P M ) or a completion-valid point (P C ) by the application of a stochastic gradient descent (or other optimizer), but does not guarantee to return the closest point, since it depends on the structure of the neural network, which is typically high-dimensional non-convnex manifold. This is different than the case in [22].…”
Section: Theoretical Analysiscontrasting
confidence: 72%
“…This section describes the condition for the convergence of the algorithm and provides some theoretical insights and in a sense the derivation is a bit similar to [22]. The convergence does not assume convexity, but it does assume certain properties for the non-linear projection operators.…”
Section: Theoretical Analysismentioning
confidence: 99%
“…B. Algoritmo IZMA_SD Gil Shabat y Amir Averbutch [21] plantean un algoritmo para la aproximación de matriz (IZMA, Interest Zone Matrix Aproximattion); uno de los métodos propuestos por los autores plantea que una matriz de bajo rango se puede aproximar al minimizar la norma de Frobenius teniendo en cuenta las restricciones dadas en la norma nuclear de la matriz.…”
Section: Soluciónunclassified
“…En Matrix Completion las entradas conocidas no deben cambiar de valor mientras se minimiza la función objetivo sujeta a las restricciones, ya que se desea minimizar el rango de la matriz. Los datos perdidos de la matriz son aquellas que no se muestrean y el objetivo es reconstruir la señal; en este trabajo se modifica un algoritmo IZMA de Shabat y Averbuch, descrito en [21], que busca minimizar la suma de valores singulares sujeto a un conjunto de restricciones; la modificación del mismo consiste en asignar a los datos perdidos valores semillas, partir de la desviación estándar (SD, Standard Deviation) de las muestras, este valor puede ser negativo o positivo, depende del valor anterior y posterior al dato perdido, a este algoritmo se llamará IZMA_SD.…”
Section: Soluciónunclassified
See 1 more Smart Citation