Speaker:Chenglin Fan

Title: In Pursuit of Metric Data and Privacy in Modern Data Analysis

Abstract: 

Modern data analysis presents many new challenges less considered by classic settings. First, many classic data analysis algorithms either assume or are considerably more efficient if the distances between the data points satisfy a metric. However, as real data sets are noisy, they often do not possess this fundamental property.  Second, data leaks and commercial data trading threaten to reveal sensitive information contained in these data. Thus, it is necessary to develop algorithms for addressing non-metric data and private data.

In this talk, I will present algorithms that rise to these aforementioned challenges. I will first formalize the  metric repair problem (which seeks to minimally modify the data to make it metric, thereby finding the nearest metric data set) and propose approximate solutions to that problem.  Then, I will present  private algorithms for several classic problems such as k-median, correlation clustering and more.

Biography:

Chenglin Fan is a postdoctoral researcher at Pennsylvania State University. He received his Ph.D. in Computer Science from UT Dallas. After graduation, he did a postdoc at Sorbonne University and subsequently was a researcher at Baidu Research. His research interests lie broadly in theoretical computer science and differential privacy. Dr. Fan's research has been published at top theory and machine learning conferences such as FOCS, SODA, ICML and NeurIPS.