I'm currently a fourth-year Ph.D. student in the Department of Mathematics, National University of Singapore.
I began my Ph.D. in August 2022 in Mathematics.
In September 2023, I shifted my research focus to Machine Learning and started working under the supervision of Professor Tan Minh Nguyen.
Previously, I completed an Honors degree in Mathematics at the Hanoi University of Science (2018-2022).
email: hoang.tranviet [at] u.nus.edu | Google Scholar
Symmetry of Parameter Space. This research direction aims to completely characterize the symmetries inherent in modern neural architectures and to translate these theoretical insights into practical methodologies, including the analysis of (linear) mode connectivity, the design of symmetry-respecting optimizers, and the development of equivariant metanetworks.
Efficient methods based on Optimal Transport (OT). Our research develops scalable OT-based techniques, with a particular focus on tree-sliced OT, a generalization of sliced OT that exploits closed-form solutions of OT problems on tree metric spaces.
keywords: machine learning, parameter space symmetry, mode connectivity, metanetwork, optimal transport
Jan 2026. 4 papers Quasi-Equivariant Metanetworks, Mixed-Curvature Tree-Sliced Wasserstein Distance, Tree-sliced Sobolev IPM, and Revisiting Tree-Sliced Wasserstein Distance Through the Lens of the Fermat-Weber Problem, are accepted at ICLR 2026
Oct 2025. 1 paper Modeling Expert Interactions in Sparse Mixture of Experts via Graph Structures is accepted at AAAI 2026 Workshop MATH4AI
Sep 2025. 3 papers On Linear Mode Connectivity of Mixture-of-Experts Architectures (oral), Tree-Sliced Entropy Partial Transport, and Dynamical Properties of Tokens in Self-Attention and Effects of Positional Encoding, are accepted at NeurIPS 2025
Jun 2025. I will give a talk on Functional Equivalence in Neural Architectures, in Joint FDU-NUS-ZJU PhD Forum in Mathematics, at Department of Mathematics, National University of Singapore
May 2025. 3 papers Equivariant Polynomial Functional Networks, Tree-Sliced Wasserstein Distance: A Geometric Perspective, and Tree-Sliced Wasserstein Distance with Nonlinear Projection, are accepted at ICML 2025
>
older news
On Linear Mode Connectivity of Mixture-of-Experts Architectures
Viet-Hoang Tran¹, Van Hoan Trinh¹, Khanh Vinh Bui¹, Tan M. Nguyen²
NeurIPS 2025 (oral)
Equivariant Neural Functional Networks for Transformers
Viet-Hoang Tran¹, Thieu N. Vo¹, An Nguyen The¹, Tho Tran Huu, Minh-Khoi Nguyen-Nhat, Thanh Tran, Duy-Tung Pham, Tan Minh Nguyen²
ICLR 2025
Distance-Based Tree-Sliced Wasserstein Distance
Viet-Hoang Tran¹, Khoi N.M. Nguyen¹, Trang Pham, Thanh T. Chu, Tam Le², Tan M. Nguyen²
ICLR 2025