Studying and Improving Graph Neural Network-based Motif Estimation
Vieira, P. C., Silva, M. E. P., & Ribeiro, P. M. P. (2025). Studying and Improving Graph Neural Network-based Motif Estimation. doi:10.48550/ARXIV.2506.15709
Table of Contents
✦ ✦ ✦
Education
✦ ✦ ✦
Master in Data Science
-
Faculty of Sciences of University of Porto -
Porto, Portugal -
September 2022 - July 2024 - Final Mark: Average 19/20; ECTS grade A; 4.0/4.0 GPA (top 1%);
- Note: Thesis in viva approved with the maximum grade possible.
Bachelor in Computer Science
-
Faculty of Sciences of University of Porto
-
Porto, Portugal
-
September 2019 - July 2022
- Final Mark: Average 18/20; ECTS grade A; 3.90/4.0 GPA (top 1%)
Summer School for Introduction to Artificial Intelligence
-
Université Paris-Saclay à Centrale Supélec
-
Orsay, Paris, France
-
June 2022 - July 2022
- Final Mark: Pass
Reseach Experience
✦ ✦ ✦
1.. Research Assistant as Master’s Candidate
-
Department of Computer Science of the Faculty of Sciences of University of Porto -
Porto, Portugal
-
September 2023 - July 2024
Short Description:
- Conducted independent research under the mentorship of Prof. Pedro Manuel Pinto Ribeiro and Prof. Miguel Eduardo Pinto da Silva, with a focus on advancing Graph Neural Networks (GNNs) for motif discovery.
- Explored the foundational principles, limitations, and current methodologies of GNNs, critically analyzing state-of-the-art techniques and identifying significant gaps in the literature.
- Designed and executed both theoretical and empirical experiments addressing these gaps, culminating in an alternative to motif estimation using GNNs.
- Led the drafting of a manuscript detailing these findings for publication.
2.. Research Assistant
-
INESC TEC - Artificial Intelligence and Decision Support Research Center
-
Porto, Portugal
-
February 2022 - January 2023
Short Description:
- Collaboration between Carnegie Mellon (CMU), INESC TEC/ID, NOVA.ID.FCT and FCiências.ID to deanonymize Dark Web Traffic of Mix Networks for Cybercrime Investigation (project: DAnon)
- Applied machine learning methods, including reinforcement learning and genetic algorithms, to extract origin relay information from encrypted TCP connections.
- Developed and implemented heuristic-based algorithms for temporal sequence similarity measurement to enhance relay matching accuracy.
- Researched the applicability of the described methods to demultiplex encrypted TCP requests from a single logical connection.
- Authored a comprehensive technical report documenting methodologies, findings, and implications for secure communications analysis (not publicly available).
3.. Research Assistant
-
INESC TEC - HumanISE Research Center
-
Porto, Portugal
-
February 2021 - September 2021
Short Description:
- Collaborated with a research team (MoST project) to design advanced spatio-temporal data structures, integrating rigorous time and space complexity analysis to establish defined performance boundaries, beating the state-of-the-art.
- Explored structural applications to challenging spatio-temporal problems, notably the “Teapot in a Stadium” problem, deriving critical insights that contributed to error correction and refinement of the models.
- Led the drafting of a manuscript detailing these findings for publication.
Publications
✦ ✦ ✦
-
The Dynamic-Loose octree : a spatial index structure towards time-efficiency
Carneiro, E., C. Vieira, P., Carvalho, A. V., Amaro, M., The Dynamic-Loose octree : a spatial index structure towards time-efficiency.
-
S+t-SNE - Bringing Dimensionality Reduction to Data Streams
C. Vieira, P., Montrezol, J. P., T. Vieira, J., & Gama, J. (2024). S+t-SNE - Bringing Dimensionality Reduction to Data Streams. In I. Miliou, N. Piatkowski, & P. Papapetrou (Eds.), Advances in Intelligent Data Analysis XXII (pp. 95–106). Cham: Springer Nature Switzerland.
Talks
✦ ✦ ✦
-
From Networks to Insights: Analyzing Graphs With and Without Machine Learning
Invited PhD Interview Talk at The International Max Planck Research School for Intelligent Systems, Stuttgart, Germany
-
S+t-SNE Bringing dimensionality reduction to data streams
Talk at Intelligent Data Analysis (IDA), Stockholm, Sweden
Academic Activites and Service
✦ ✦ ✦
Reviewer
- 28th International Conference on Discovery Science 2025 - Program Committee (Reviewer)
- European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases (ECML-PKDD) 2025 - Reviewer
- International Conference on Learning Representations (ICLR) 2025 - Reviewer
- 27th International Conference on Discovery Science 2024 - Program Committee (Reviewer)
Academic Service
- Monitoring Committee of the Master Program in Data Science
-
Member -
December 2022 - December 2024
-
- Monitoring Committee of the Bachelors Program in Computer Science
-
Member
-
December 2021 - December 2022
-
Selected Extra Projects
✦ ✦ ✦
1.. Construction and Analysis of Complex Networks of MOBA Matches
- Worked in retrieving massive data regarding MOBA matches using public APIs and in the construction of complex networks from such data.
- Examined the created networks regarding properties such as bond and site percolation and information diffusion.
- Applied machine learning algorithms in order to predict the rank of a match.
2.. Maze Solver
- Explored computer vision techniques to digitally reconstruct mazes given by live video.
- Examined different methodologies to make the reconstruction resilient to different types of illumination and different ways of representing the maze e.g. on a piece of paper and on backlit screens.
- Implemented simple algorithmic tricks to solve the digital reconstruction.
3.. Modulation and Forecast of Aerosol Time Series
- Studied time series regarding aerosol quantities in different points of the world and modulated their behavior with multiple parametric and non-parametric models.
- Compared different methods for forecasting, together with a study of best practices for forecasting.
4.. Random Polygon Generator
- Designed algorithms based on adversarial and local search and optimisation to efficiently generate convex polygons given a set of points and predetermined connections.
- Investigated shortest route discovering (TSP-like) methodologies through the aforementioned method.
Skills and Interests
✦ ✦ ✦
Technical
-
Programming Languages:- Expert: Python, Java, R
- Advanced: Bash, C++, C, Haskell, MySQL
- Beginner: CUDA, Rust, Erlang
-
Tools and Frameworks:- Expert: Pytorch, Pytorch Geometric, TensorFlow/Keras, Numpy, Pandas, Scikit-Learn, Networkx, Networkit, Graph-Tool, Ray
- Intermediary: OpenCV, Weka, Gephi, AMPL, BLAST, biopython, PySpark
- Beginner: SLURM, OpenSearch, RAPIDS, HoloViz
Language
-
Native or Native Level: Portuguese, English -
Reading Proficiency: Spanish
Other Interests
- Building computers and experimenting with hardware; Studying mythology (mainly greek
)
References
✦ ✦ ✦
João Manuel Portela da Gama
Pedro Manuel Pinto Ribeiro
Miguel Eduardo Pinto da Silva
Full Professor
Associate Professor
Invited Professor; Data Scientist
University of Porto
University of Porto
University of Porto, EXADS
jgama@fep.up.pt
pribeiro@fc.up.pt
mepsilva@fc.up.pt