Signal Analysis for Improved Machine Learning
Abstract
This project will focus on a fundamental research issue pertaining to signal processing algorithms for improved machine learning in big-data problems. In particular, the study will undertake investigations on simultaneous segmentation and clustering scheme by means of Non-negative Matrix Factorization or similar feature mapping. However, there are too few dataset with known labels that can be used to train for the clustered categories. To mitigate this challenge, an effective semi-supervised training will be explored so that a small labeled dataset can be propagated into the unlabeled ones and transform them into highly accurate labeled dataset sufficient for achieving generalization. Hence, the goal is to address the inadequate amount of labeled training dataset by investigating for an effective semi-supervised learning so that high-quality training data can be made available for training generalization.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2016
- Source ID
- N000141512428
Entities
People
- Hanseok Ko
Organizations
- Korea University
- Office of Naval Research
- United States Navy