劉智弘 Chih-Hung Liu

Associate Professor, Department of Electrical Engineering, National Taiwan University

Major Research Areas:

Computational Geometry, Randomized Algorithms, Algorithmic Aspects of Data Science, Computer Vision

Research Summary:

News

• STOC 2024: Private Graphon Estimation via Sum-of-Squares.
  • with Hongjie Chen, Jingqiu Ding, Tommaso d'Orsi, Yiding Hua and David Steurer
• NeurIPS 2024: Robust Sparse Regression with Non-Isotropic Designs.
  •  with Gleb Novikov

 

--- Algorithmic Aspects of Data Science

Algorithmic Aspects of Data Science represent a rapidly growing area in Theoretical Computer Science and Data Science. Over the past two decades, Data Science has sparked a technological revolution across a wide range of scientific and industrial fields. Despite its widespread application, the theoretical foundations underlying its success remain partially unexplored. In this context, the Algorithmic Aspects of Data Science aim to investigate fundamental problems in Data Science from a theoretical perspective, focusing on the development of efficient algorithms with rigorous theoretical guarantees. My primary research topics in this area include Tensor Problems, Robust Statistics, and Clustering.

--- Computer Vision and Machine Learning (the original team by Prof. Sy-Yen Kuo)

This topic focuses on developing algorithms that enable machines to observe, analyze, and understand visual information, as well as learn from data to make predictions. Our research spans a broad spectrum of applications, including image recognition, object detection and tracking, the enhancement of visual perception, and Generative AI. We have many research achievements published in top conferences such as CVPR, ECCV, ICCV, and AAAI.

--- Computational Geometry

Computational Geometry is a fundamental area in Theoretical Computer Science, where we investigate the combinatorial properties of geometric problems and develop efficient algorithms. I am focused on proximity problems, involving distances between geometric objects. For example, I have published many results for Voronoi diagrams. Voronoi diagrams lie at the heart of computational geometry and serve as an extremely important tools for proximity problems. Furthermore, I am also studying the k nearest neighbors problem, which have many applications in diverse areas including data science.