Roberto Gheda

PhD Student at TU Delft, Machine Learning and Network Science

About Me

I am Roberto Gheda, PhD student at TU Delft. My research interest spans across graph theory, security and fairness in machine learning, and bayesian networks. I have been working along companies such as ASML and KPN. Feel free to reach out to me for any collaboration opportunity!

Publications

First author publications:

  • “Checkmate! Watermarking Graph Diffusion Models in Polynomial Time”.
    Roberto Gheda, Abele Malan, Robert Birke, Maksim Kitsak, Lydia Chen.
    TL;DR First to apply watermark on graph data in polynomial time by using graph diffusion models and spectra of random matrices.
    ICLR 2026. [Paper]
  • “Collaborative and Confidential Junction Trees for Hybrid Bayesian Networks”.
    Roberto Gheda, Abele Malan, Thiago Guzella, Carlo Lancia, Robert Birke, Lydia Chen.
    TL;DR Multi-party root-cause analysis framework for Bayesian Networks, with improved scalability and precision.
    NeurIPS 2025. [Paper] [Poster]

Other publications:

  • “Test-time Graph Extrapolation via Progressive Anchor-Guided Expansion”
    Abele Malan, Roberto Gheda, Robert Birke, Lydia Chen.
    TL;DR Sampling conditioning to preserve validity of graphs generated by diffusion models at scale.
    ICLR (TTU Workshop) 2026. [Paper]
  • “SuperHype: Hypergraph Generation via Graph-Superposition Decomposition”.
    Lucas Gantes, Abele Malan, Roberto Gheda, Robert Birke, Lydia Chen.
    TL;DR A diffusion model for hypergraphs, with a scalable latent representation.
    Under review, 2026.
  • “MAGiC: Attributed Graph Generation via Mixed-type Diffusion and Coarsening”.
    Abele Malan, Roberto Gheda, Robert Birke, Lydia Chen.
    TL;DR A diffusion model for generating graphs with rich node features.
    Under review, 2026.

Education

PhD, TU Delft

03/2025 - now

PhD Student in Machine Learning and Network Science.

Advisors: Prof. Lydia Chen, Prof. Maksim Kitsak

MSc, TU Delft

09/2022 - 01/2025

MSc Machine Learning.

Thesis: Collaborative and Confidential Bayesian Networks.

Advisors: Prof. Lydia Chen, Dr. Thiago Guzella

BSc, University of Trento

09/2019 - 07/2022

BSc Computer Science.

Thesis: Integration of PlanSys2 for a fleet of AGVs.

Advisor: Prof. Marco Roveri

Industry Collaboration

KPN

03/2025 – present

Secure Data Sharing for Networks.

ASML

03/2024 – 01/2025

Collaborative and confidential root-cause analysis for semiconductor manufacturing via Bayesian Networks.

References