Huyen Vo (Emily)

Hi, I’m Huyen. You can also call me Emily. 🍀

I’m a PhD student at Saarland University (UdS) and Max Planck Institute for Software Systems (MPI-SWS), part of the CS@Max Planck PhD program, where I’m fortunate to be advised by Prof. Isabel Valera in the Probabilistic Machine Learning group.

👩‍💻 My research focuses on multimodal generative models, with a particular interest in the optimization and inference challenges that arise when learning joint representations across multiple data modalities. Previously, I worked on implicit models such as neural ODEs. Across my work, I aim to leverage probabilistic and optimization perspectives to develop models that are both effective and efficient.

Before my PhD, I was a Research Resident at FPT Software AI Center, working with Prof. Tan Minh Nguyen and Dr. Thieu N. Vo. In 2023, I graduated from the Honors Program at Hanoi University of Science and Technology (HUST) with a Computer Science Bachelor’s degree. During my time there, I was a research student at the Data Science Laboratory, advised by Dr. Linh Van Ngo.

I’m always happy to chat about new and interesting research ideas - feel free to reach out! 🤗

Besides research, I have an infinite love for mountains and hiking — the Alps have my whole heart. 🏔️

news

Jan 22, 2026	My first PhD paper Hellinger Multimodal Variational Autoencoders got a Spotlight (2.5%) at AISTATS 2026! 🎉
Oct 01, 2024	I started my PhD at the CS@Max Planck program. Excited for the journey ahead! 👩‍💻
Sep 15, 2022	My first paper Improving Neural Ordinary Differential Equations with Nesterov’s Accelerated Gradient Method got accepted at NeurIPS 2022!

selected publications

AISTATS
Spotlight

Hellinger Multimodal Variational Autoencoders

Huyen Vo and Isabel Valera

In The 29th International Conference on Artificial Intelligence and Statistics, 2026

Spotlight Presentation [2.5%]

Abs PDF Code

Spotlight

Multimodal variational autoencoders (VAEs) are widely used for weakly supervised generative learning with multiple modalities. Predominant methods aggregate unimodal inference distributions using either a product of experts (PoE), a mixture of experts (MoE), or their combinations to approximate the joint posterior. In this work, we revisit multimodal inference through the lens of probabilistic opinion pooling, an optimization-based approach. We start from Hölder pooling with α=0.5, which corresponds to the unique symmetric member of the α-divergence family, and derive a moment-matching approximation, termed Hellinger. We then leverage such an approximation to propose HELVAE, a multimodal VAE that avoids sub-sampling, yielding an efficient yet effective model that: (i) learns more expressive latent representations as additional modalities are observed; and (ii) empirically achieves better trade-offs between generative coherence and quality, outperforming state-of-the-art multimodal VAE models.
NeurIPS

Improving Neural Ordinary Differential Equations with Nesterov’s Accelerated Gradient Method

Nghia Nguyen, Tan Minh Nguyen, Huyen Vo, and 2 more authors

In Advances in Neural Information Processing Systems, 2022

Abs PDF Code

We propose the Nesterov neural ordinary differential equations (NesterovNODEs), whose layers solve the second-order ordinary differential equations (ODEs) limit of Nesterov’s accelerated gradient (NAG) method, and a generalization called GNesterovNODEs. Taking the advantage of the convergence rate O(1/k^2) of the NAG scheme, GNesterovNODEs speed up training and inference by reducing the number of function evaluations (NFEs) needed to solve the ODEs. We also prove that the adjoint state of a GNesterovNODEs also satisfies a GNesterovNODEs, thus accelerating both forward and backward ODE solvers and allowing the model to be scaled up for large-scale tasks. We empirically corroborate the advantage of GNesterovNODEs on a wide range of practical applications, including point cloud separation, image classification, and sequence modeling. Compared to NODEs, GNesterovNODEs require a significantly smaller number of NFEs while achieving better accuracy across our experiments.