Visiting address

DS - BI
University of Groningen
Nijenborgh 9
9747 AG Groningen
The Netherlands

Quick links

Follow us

TODO

Huy Truong

An Online, Continuous, Self-Adaptive Pipeline for Water Distribution Network State Estimation

Supervisors: Huy Truong, Dilek Düştegör
Date: 2026-01-09
Type: bachelor-project/master-project/master-internship
Description:

Deploying machine-learning pipelines in real-world systems is challenging due to data distribution drift and inherent instability. Conventional machine-learning models typically rely on fixed weights optimized for a specific training distribution, which leads to performance degradation when exposed to unseen and noisy data in practice. To address this limitation, this group project develops a framework that supports online learning and introduces an evaluation benchmark that more closely reflects real-world operating conditions. Specifically, the project consists of two main components: 1. A pipeline built around a pretrained Graph Neural Network (GNN) to estimate unknown hydraulic measurements from a limited set of sensors deployed across a water distribution network. This component focuses on implementing a Test-Time Training strategy that adapts model weights using only incoming test inputs. 2. A benchmarking platform that simulates real-world steady-state snapshots, incorporating hydraulic measurements such as pressure, demand, and network topology across multiple water distribution systems. The benchmark is designed to evaluate the robustness and adaptability of machine-learning pipelines under what-if analyses and out-of-distribution conditions. References:
Test-Time Training with Self-Supervision for Generalization under Distribution Shifts.

Multivariate State Estimation in Drinking Water Distribution Networks

Supervisors: Huy Truong, Andrés Tello, Alexander Lazovik
Date: 2026-01-09
Type: bachelor-project/master-project/master-internship
Description:

Monitoring water distribution networks plays the main role in ensuring safe drinking water delivery to millions of residents in the urban area. Traditionally, this task relies on physics-based mathematical simulations; however, such models require a large number of parameters and frequent recalibration to maintain accuracy consistent with sensor measurements. As an alternative, recent studies have proposed data-driven approaches based on Graph Neural Networks (GNNs), which leverage pressure measurements from a limited set of sensors at known locations to infer pressure values at unmonitored nodes in the network. Building on this idea, the project extends the existing univariate method to a multivariate framework, aiming to jointly estimate multiple hydraulic quantities, including pressure, demand, flow rate, head loss, and others. The candidate is expected to have a basis of machine-learning foundation and proficiency in one of the deep learning frameworks (PyTorch, TensorFlow). Reference:
Graph Neural Networks for Pressure Estimation in Water Distribution Systems**.**

Graph Reasoning Models

Supervisors: Huy Truong, Dilek Düştegör
Date: 2026-01-09
Type: master-project/master-internship
Description:

Graph Neural Networks (GNNs) have emerged as promising approaches in processing graph-based systems. GNNs leverage a message passing mechanism to update node features given neighborhood information. However, this mechanism often paired with several issues, particularly for over-smoothing, a phenomenon in which GNNs encode similar representations for all nodes in the graph. This hinders the scalability, constraining these models’ depth to a shallow level. This work explores a recursive approach to extend the number of layers virtually while measuring the impact of over-smoothing in this specific setting. The new approach is validated in the context of the water domain. Students interested in joining this project should have a basis of machine-learning knowledge and be familiar with one of deep-learning frameworks (PyTorch, Tensorflow). References:
Less is More: Recursive Reasoning with Tiny Networks. Assessing the performances and transferability of graph neural network metamodels for water distribution systems**.** Hierarchical Reasoning Model**.**

Dataset Management

Supervisors: Huy Truong, Andrés Tello
Date: 2025-08-25
Type: bachelor-internship
Description:

Have you ever managed a large dataset? This project provides an opportunity to handle a dataset with over 8,000 downloads each month. You will reorganize the dataset by task, document it thoroughly, and create a user-friendly interface and leaderboard. The project also involves working with HPC clusters, Hugging Face libraries, and GitHub Pages for documentation. Basic Python skills and familiarity with Linux commands are required.

Node masking in Graph Neural Networks

Supervisors: Huy Truong, Dilek Düştegör
Date: 2025-01-21
Type: bachelor-project
Description:

Working with data in the real world often leads to missing information problems, which can negatively affect the performance of deep learning models. However, in proper ways, it can boost the expressiveness of Graph Neural Network (GNN) models in node representation learning through a technique known as Node Masking. In particular, it hides arbitrary nodal features in a graph and instructs the GNN to recover the missing parts. The student can explore diverse masking strategies, such as zero masking, random node replacement, mean-neighbor substitution, shared learnable embedding, and nodal permutation. These options above should be compared and evaluated in a graph reconstruction task that applies to a water distribution network. This study will focus on finding a generative technique that effectively enhances the performance of GNN models in semi-supervised transductive learning. Students interested in joining this project should possess a machine-learning background and a deep-learning framework.
References:
Hou, Zhenyu, Xiao Liu, Yuxiao Dong, Chunjie Wang, and Jie Tang. "GraphMAE: Self-Supervised Masked Graph Autoencoders." arXiv preprint arXiv:2205.10803(2022). Abboud, Ralph, Ismail Ilkan Ceylan, Martin Grohe, and Thomas Lukasiewicz. "The surprising power of graph neural networks with random node initialization." arXiv preprint arXiv:2010.01179 (2020). Hajgató, Gergely, Bálint Gyires-Tóth, and György Paál. "Reconstructing nodal pressures in water distribution systems with graph neural networks." arXiv preprint arXiv:2104.13619 (2021). He, Kaiming, Xinlei Chen, Saining Xie, Yanghao Li, Piotr Dollár, and Ross Girshick. "Masked autoencoders are scalable vision learners." In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 16000-16009. 2022.

Multimodality Graph Foundation Models

Supervisors: Huy Truong
Date: 2025-01-21
Type: student-colloquium
Description:

References:
Lam, Hoang Thanh, et al. "Otter-Knowledge: benchmarks of multimodal knowledge graph representation learning from different sources for drug discovery." *arXiv preprint arXiv:2306.12802* (2023). Xia, Lianghao, Ben Kao, and Chao Huang. "Opengraph: Towards open graph foundation models." *arXiv preprint arXiv:2403.01121* (2024). Ektefaie, Yasha, et al. "Multimodal learning with graphs." *Nature Machine Intelligence* 5.4 (2023): 340-350.

Test-Time Training

Supervisors: Huy Truong
Date: 2025-01-21
Type: student-colloquium
Description:

References:
Sun, Yu, et al. "Test-time training with self-supervision for generalization under distribution shifts." International conference on machine learning. PMLR, 2020. Liu, Yuejiang, et al. "Ttt++: When does self-supervised test-time training fail or thrive?." Advances in Neural Information Processing Systems 34 (2021): 21808-21820. Liang, Jian, Ran He, and Tieniu Tan. "A comprehensive survey on test-time adaptation under distribution shifts." *International Journal of Computer Vision* (2024): 1-34. Behrouz, Ali, Peilin Zhong, and Vahab Mirrokni. "Titans: Learning to Memorize at Test Time." arXiv preprint arXiv:2501.00663 (2024).

DiTEC project- Inverse problem in Water Distribution Networks

Supervisors: Huy Truong
Date: 2025-04-03
Type: master-internship
Description:

Water researchers have relied on simulations to monitor the behavior of Water Distribution Networks. These simulations require a comprehensive set of parameters- such as elevation, demand, and pipe diameters-to determine hydraulic states accurately. This increases labor cost, time consumption and, therefore, poses a significant challenge. But what if we could reverse the process and let AI infer the missing pieces? Building on this idea, the project explores an innovative approach: leveraging data-driven deep learning methods to predict initial input conditions based on available output states. As a researcher on this project, the candidate will select and train a cutting-edge Graph Neural Network on a massive dataset. As a result, the model should be able to predict initial conditions while considering the structural and physical constraints. The candidate will submit the implementation code and a report detailing the problem and the proposed solution. The ideal candidate should have a background in machine learning and be familiar with at least one deep-learning framework.
References:
Truong, Huy, et al. "DiTEC-WDN: A Large-Scale Dataset of Hydraulic Scenarios across Multiple Water Distribution Networks." (2025).

Can we train a Neural Network with Forward-Forward “harmoniously”?

Supervisors: Huy Truong
Date: 2025-04-03
Type: master-internship
Description:

Back Propagation(BP) is a de facto approach to training neural network models. Nevertheless, it is biologically implausible and requires complete knowledge (i.e., tracks the entire flow of information from start to end of a model) to perform a backward pass. Instead, an alternative approach called Forward-Forward(FF) can replace the backward pass with an additional forward one and update the model weights in an unsupervised fashion. In particular, FF performs forward passes using positive and negative inputs, respectively, and, therefore, employs the difference between the two activation versions at each layer in the neural network to compute the loss and update weights. Here, the project studies the behavior of FF employing different losses: (1) cross-entropy and (2) harmonic loss. Also, it is valuable to study the relevance between harmonic loss and FF in terms of distance metrics or geometric properties in an embedding space. As a deliverable, the candidate should submit a detailed report and implementation code. For primary requirements, the candidate should be familiar with one of the deep learning frameworks and have experience in setting up machine learning experiments.
References:
Hinton, Geoffrey. "The forward-forward algorithm: Some preliminary investigations." *arXiv preprint arXiv:2212.13345* (2022). Baek, David D., et al. "Harmonic Loss Trains Interpretable AI Models." *arXiv preprint arXiv:2502.01628* (2025).