Communications of the ACM
Thinking Outside the Sample
A*STAR researchers have presented a new machine learning framework to solve big-data problems.
Credit: R. Pavel/Getty
Researchers at Singapore's Agency for Science, Technology, and Research (A*STAR) have developed a framework they say could help computers learn how to process and identify images faster and more accurately.
The framework can be used for numerous applications, including image segmentation, motion segmentation, data clustering, hybrid system identification, and image representation, according to A*STAR researcher Peng Xi.
Conventional computers process data using representation learning, which involves identifying a feature that enables the program to quickly extract relevant information from the dataset and categorize it. Supervised and unsupervised learning are two of the main methods used in representation learning. Supervised learning relies on costly labeling of data prior to processing, while unsupervised learning involves grouping or "clustering" data in a similar manner to human brains, according to Peng.
Subspace clustering is a form of unsupervised learning that aims to fit each data point into a low-dimensional subspace to find an intrinsic simplicity that makes complex, real-world data tractable.
"By solving the large-scale data and out-of-sample clustering problems, our method makes big-data clustering and online learning possible," Peng says.
The researchers tested their new method on a range of datasets and found the framework outperformed existing algorithms and successfully reduced the computational complexity of the task while still ensuring cluster quality.
From A*STAR Research
View Full Article
Abstracts Copyright © 2016 Information Inc., Bethesda, Maryland, USA
No entries found