Akang Wang

Operations Research

prof_pic.png

I’m Akang Wang, a Research Associate at the Shenzhen Research Institute of Big Data and an Adjunct Assistant Professor in the School of Data Science at CUHK-Shenzhen. I received my PhD in Process Systems Engineering from Carnegie Mellon University in 2020. My research centers on mathematical optimization, spanning theory, algorithms, and applications. I am currently interested in inference optimization for LLMs. My research is supported by National Natural Science Foundation of China, National Key R&D Program of China, and Guangdong Basic and Applied Basic Research Foundation. My work has appeared at leading machine learning conferences, including ICLR, ICML, and NeurIPS. I won first place in the primal track of the NeurIPS 2021 ML2CO Competition and second place in the 2022 INFORMS RAS Problem Solving Competition. I also serve as a reviewer for venues including NeurIPS/ICML/ICLR.

In this website, you can know more about my research and experience. Also, please feel free to check out my Google Scholar and LinkedIn.

news

Aug 03, 2026 Our algorithm PDBO has produced the SOTA results for the challenging LABS instances from QOBLIB.
Jul 18, 2026 Our paper entitled “An Accelerated Mixed Weighted-Unweighted MMSE Approach for MU-MIMO Beamforming” is accepted by IEEE Transactions on Signal Processing.
May 16, 2026 I gave an invited talk entitled “Smoothing Binary Optimization: A Primal-Dual Perspective” (slides) at The 3nd Youth Forum of Mathematical Programming Branch of the Operations Research Society of China.
Mar 02, 2026 Our paper entitled “Parallel Graver Basis Extraction for Nonlinear Integer Optimization” is accepted by Operations Research Letters.
Jan 28, 2026 Our paper entitled “Successive Fixing for Large-Scale SCUC Using First-Order Methods” is accepted by PSCC 2026.
Nov 22, 2025 Our paper entitled “The Periodic Vehicle Routing Problem with Multi-Day Trips” is accepted by Transportation Research Part E.
Nov 19, 2025 Our paper entitled “On representing convex quadratically constrained quadratic programs via graph neural networks” is accepted by TMLR.
Sep 29, 2025 We are happy to share that a variant of our algorithm PDBO produced a new incumbent for QPLIB_2880.

latest posts