Alvin Heng

Hello! I am currently a PhD student at the National University of Singapore, where I work on deep generative models as part of CLeAR Lab with Harold Soh.

Research: The goal of my research is to make generative models ready for the real-world, by focusing on improving three aspects: safety, robustness and efficiency. As part of my PhD, some areas that I have worked on include gradient flows, diffusion models, variational inference and time-series prediction. To learn more about my background and research, please refer to my CV and Google Scholar profile linked below.

Previously: In my past life, I worked as a physicist, specializing in the field of condensed matter physics during my undergraduate studies. My primary focus involved conducting numerical simulations to study exotic physical phenomena in quantum materials. During this time, I worked with Pinaki Sengupta at the Nanyang Technological University, as well as Anna Keselman and Leon Balents at the Kavli Institute for Theoretical Physics at UC Santa Barbara. As a Masters student at the University of Toronto, I worked with Nathan Wiebe to explore how classical deep learning techniques can be used to improve the performance of quantum algorithms.

Email  /  CV  /  Google Scholar  /  GitHub  /  LinkedIn

profile photo

Publications

project image

Out-of-Distribution Detection with a Single Unconditional Diffusion Model


Alvin Heng, Alexandre H. Thiery, Harold Soh
Advances Neural Information Processing Systems 37 (NeurIPS), 2024
arxiv / code

project image

Selective Amnesia: A Continual Learning Approach to Forgetting in Deep Generative Models


Alvin Heng, Harold Soh
Advances Neural Information Processing Systems 36 (NeurIPS), 2023, Spotlight
arxiv / code

project image

Neural Continuous-Discrete State Space Models for Irregularly-Sampled Time Series


Abdul Fatir Ansari, Alvin Heng, Andre Lim, Harold Soh
International Conference on Machine Learning (ICML), 2023, Oral
arxiv / code

project image

Generative Modeling with Flow-Guided Density Ratio Learning


Alvin Heng, Abdul Fatir Ansari, Harold Soh
Joint European Conference on Machine Learning and Knowledge Discovery in Databases (ECML PKDD), 2024
arxiv / code

project image

Three-Magnon Bound State in the Quasi-One-Dimensional Antiferromagnet a-NaMnO2


Rebecca L Dally, Alvin Heng, Anna Keselman, Mitchell M Bordelon, Matthew B Stone, Leon Balents, Stephen D Wilson
Physical Review Letters
arxiv / paper

project image

Pair Hopping in Systems of Strongly Interacting Hard-Core Bosons


Alvin Heng, Wenan Guo, Anders W Sandvik, Pinaki Sengupta
Physical Review B
arxiv / paper

project image

Optimal Fee Structure For Efficient Lightning Networks


Alvin Heng, Ling Feng, Siew Ann Cheong, Rick Siow Mong Goh
IEEE 24th International Conference on Parallel and Distributed Systems (ICPADS), 2018
paper


Website credits: 1, 2