Posted May 24, 2019
Data Recovery and Subspace Clustering from Quantized and Corrupted Measurements
Team Members and Titles
Summary

New subclustering methods to recover data from degraded or corrupted data to improve image and video processing, network analysis, and data privacy applications.

Key Findings

Algorithm to accurately recover the original high-resolution data and cluster data accurately from low quality data.

In this project, we addressed data clustering, one fundamental machine learning problem that groups data points into clusters based on their similarities. We, for the first time, developed a data clustering methods with provable guarantees to recover and cluster data simultaneously from highly quantized measurements. Our research shows that even if the data quality is extremely low, as long as large amounts of data are available, one can still accurately recover the original high-resolution data and cluster the data points accurately. This insight is new and advances state of the art. Our approach applies to image and video processing, network analysis, and data privacy enhancement.

 

Project Start Date
Project End Date

Publications

Ren Wang, Meng Wang, Jinjun Xiong, “Tensor recovery from quantized measurements.” submitted to The Thirty-Fourth AAAI Conference on Artificial Intelligence (AAAI) 2020, under review.

Ren Wang, Meng Wang, Jinjun Xiong, “Achieve Data Privacy and Clustering Accuracy Simultaneously Through Quantized Data Recovery,” submitted to  EURASIP Journal on Advances in Signal Processing, October 2019, (under review)

Ren Wang, Meng Wang, and Jinjun Xiong. "Data Recovery and Subspace Clustering from Quantized and Corrupted Measurements." IEEE Journal of Selected Topics in Signal Processing, Special Issue on Robust Subspace Learning and Tracking: Theory, Algorithms, and Applications, 2018, 12(6): 1547-1560.