VGU RESEARCH REPOSITORY
Please use this identifier to cite or link to this item:
https://epub.vgu.edu.vn/handle/dlibvgu/584| DC Field | Value | Language |
|---|---|---|
| dc.contributor.author | Nguyen Duc Dang Quang | en_US |
| dc.date.accessioned | 2020-03-12T17:04:32Z | - |
| dc.date.available | 2020-03-12T17:04:32Z | - |
| dc.date.issued | 2017 | - |
| dc.identifier.uri | http://epub.vgu.edu.vn/handle/dlibvgu/584 | - |
| dc.description.abstract | Matrix completion problem emerges in the scenario that there is a demand to complete a matrix when only few of its entries are exposed. Especially. the completed matrix is preferred to be a low-rank or approximately low-rank matrix. The situation is encountered in many research fields such as computer vision, machine learning. image analysis and so on. The common approach for settling this matrix completion problem is the nuclear-norm minimization. A method based on this approach can solve the matrix completion efficiently is the Iterated Soft Thresholding algorithm. The proposed algorithm would be tested and evaluated its efficiency when applied to complete 2-D gray-scale images. Another approach namely matrix factorization is also proven to complete matrices quickly and precisely. PowerFactorization method is known as a matrix factorization-based method that has been proven to solve the matrix completion and matrix recovery problem faster than the usual nuclear-norm minimization approach. In fMRI which is a technology that allows to observe neuronal activity in human brain based on the nuclear magnetic resonance principle and blood oxygen level dependent contrast, there is a desire to understand the relationship among brain functional areas in presence of stimulus. Many studies of brain connectivity often require powerful statistical analysis tools to extract meaningful functional components lilted to some specific brain activity and therefore make inference about hypotheses or predict psychological states. Data-driven independent component analysis is served as the conventional approach for functional network decomposition. In this work. the PowerFactorization method motivated from the matrix completion problem is nominated as an alternative method for brain network decomposition. As final part of this research. the PowerFactorization method is tested and evaluated especially for its ability to interpret real fMRI data to obtain meaningful brain network components corresponding to actual neuronal activity. | en_US |
| dc.language.iso | en | en_US |
| dc.publisher | Vietnamese-German University | en_US |
| dc.rights | Attribution-NonCommercial 4.0 International | * |
| dc.rights.uri | https://creativecommons.org/licenses/by-nc/4.0/ | * |
| dc.subject | Matrix completion | en_US |
| dc.subject | Power factorization | en_US |
| dc.title | Matrix completion - theory and practice | en_US |
| dc.type | Thesis | en_US |
| item.fulltext | With Fulltext | - |
| item.languageiso639-1 | other | - |
| item.grantfulltext | restricted | - |
| Appears in Collections: | Electrical Engineering and Information Technology (EEIT) | |
Files in This Item:
| File | Description | Size | Format | Existing users please Login |
|---|---|---|---|---|
| Matrix completion - theory and practice.pdf | 288.57 MB | Adobe PDF |
Page view(s)
60
checked on Aug 23, 2025
Download(s)
18
checked on Aug 23, 2025
Google ScholarTM
Check
This item is licensed under a Creative Commons License