Davide Gallon

PhD student

About Me

I am a PhD researcher in Applied Mathematics at the University of Münster and a member of Mathematics Münster. I work on diffusion models for scientific computing, neural methods for PDEs, and stochastic optimization.

Research

University of Muenster - Cluster of Excellence Mathematics

PhD in Applied Mathematics

July 2023 - Current

University profile

The PhD focuses on exploring novel techniques in Machine Learning and their mathematical understanding. Supervisor: Arnulf Jentzen.

ETH Zürich

Visiting PhD Student

March 2025–March 2026

I spent a year at ETH Zürich under the supervision of Patrick Cheridito.

Published Papers

  • Jean-Loup Dupret, Davide Gallon, Patrick Cheridito. INEUS: Iterative Neural Solver for High-Dimensional PIDEs. Accepted at NeurIPS 2026.[ArXiv] (2026). [Code]
  • Davide Gallon, Philippe von Wurstemberger, Patrick Cheridito, Arnulf Jentzen. Physics-Informed Diffusion Models in Spectral Space. ICML (2026). [Proceedings] [ArXiv] [Code]
  • Julian Kranz, Davide Gallon, Steffen Dereich, Arnulf Jentzen. SAD Neural Networks: Divergent Gradient Flows and Asymptotic Optimality via o-minimal Structures. NeurIPS (2025). [Proceedings] [Code]

Preprints

  • Davide Gallon, Arnulf Jentzen. Strong error analysis for the stochastic momentum optimizer. [ArXiv] (2026).
  • Davide Gallon, Arnulf Jentzen, Philippe von Wurstemberger. An overview of diffusion models for generative artificial intelligence. [ArXiv] (2024).
  • Davide Gallon, Arnulf Jentzen, Felix Lindner. Blow up phenomena for gradient descent optimization methods in the training of artificial neural networks. [ArXiv] (2022).

Selected Research Code

PISD — ICML 2026, first author. I wrote the PyTorch implementation for PISD, a spectral diffusion method for generating PDE solutions and reconstructing them from sparse observations. The release includes spectral preprocessing, distributed training with torchrun, inference configurations, and pretrained checkpoints. Code · Paper

INEUS — NeurIPS 2026, coauthor. I implemented part of the numerical experiments for INEUS and prepared the code for public release. Code · Experiment guide · Paper

SAD Neural Networks — NeurIPS 2025, coauthor. I wrote and maintain the experimental software, including training and evaluation for polynomial regression, the heat and Black–Scholes PDEs, and MNIST classification. Code · Paper

News

(Upcoming) I will be in Paris for NeurIPS 2026. Our paper INEUS: Iterative Neural Solver for High-Dimensional PIDEs was accepted at the conference.

(Winter semester 2026–27) I am organizing a machine learning seminar at the University of Münster.

(April 2026 - September 2026) I was the teaching assistant for the course Numerical Methods for PDEs II at the University of Münster.

(December 2025) I attended the Conference on Neural Information Processing Systems (NeurIPS 2025) in San Diego with the paper SAD Neural Networks: Divergent Gradient Flows and Asymptotic Optimality via o-minimal Structures by Julian Kranz, Davide Gallon, Steffen Dereich, and Arnulf Jentzen.

(September 2024) I attended the Mediterranean Machine Learning Summer School in Milan.

(August 2024) I attended the Deep Learning Theory Summer School at Princeton.

(April 2024 - September 2024) I was the teaching assistant for the course Numerical Optimization at the University of Münster.

Experience

Hitachi Energy

R&D Data Scientist

June 2022 - June 2023

https://www.hitachienergy.com/

Inspire the next

My primary responsibility was to design and implement innovative solutions to complex data-related problems, working closely with a international team of researchers and engineers. One of the major projects I have worked on is the development of a machine learning model to predict transformer’s performance. The resulting model was integrated into the company’s application.

Education

University of Padua

Master degree in Mathematical Engineering

October 2019 - March 2022

Graduated with 110/110 cum laude. I have dedicated my studies to numerical methods for differential equations, dynamic simulations, neural networks for deep learning, numerical methods for big data, stochastic equations.

University of Padua

Bachelor degree in Mathematics

October 2016 - September 2019

The Bachelor in Mathematics program provided me a comprehensive education in theoretical and applied mathematics.