Hubert Naguszewski

PhD Title: Memory matters : Beyond Markovian models of rare event kinetics

Supervisor: Prof David Quigley

Co-supervisor: Dr Peter Brommer

---

Background

For my undergraduate degree I studied Physics at the University of Warwick, graduating in 2022 with 1st class honours (MPhys Physics). My final year project supervisor was Dr John MarshallLink opens in a new window and the title of the project was "Classification of neutrino-induced particles, and characterisation of showers at DUNE". Poster Link opens in a new window

Undergraduate Research

During the course of my undergraduate studies, I undertook three summer research projects, each support by a Warwick Undergraduate Research Support Scheme (URSS)Link opens in a new window bursary.

Phase Transitions and Dielectric Properties of Nano-confined Water (Jul 2019 - Aug 2019)

My first year summer project supervisor was Dr Jon DuffyLink opens in a new window and the project involved operating a liquid nitrogen cooled cryostat connected to a PID temperature control system to measure the dielectric properties of a porous glass medium saturated with water.

Monte Carlo Study of Equilibrium State of NiAl Crystal (Jul 2020 - Aug 2020)

My second year summer project supervisor was Dr Peter BrommerLink opens in a new window and the project involved using Python alongside LAMMPS to generate data to investigate transition bands within NiAl.

Deformation of Droplets at High Speed (Jul 2021 - Aug 2021)

My third year summer project supervisor was Dr Radu CimpeanuLink opens in a new window and involved solving the Navier-Stokes equation for a Poiseuille flow with Basilisk implemented in C. The project involved regular meetings with a computational fluid dynamics group to discuss progress and techniques.

---

Research Interests

I am entering the fourth year of the HetSys programme, working under the supervision of Prof. David Quigley. My research focuses on using machine learning to study rare transitions in physical systems, specifically predicting the committor for the 2D Ising model using GPU-accelerated simulations with the GPU 2DIsingLink opens in a new window package. Convolutional neural networks are being explored for this purpose, and future work will investigate spatially invariant and spatial transformer networks. Further expansions include incorporating Boltzmann generators and adversarial networks to create reinforced learning loops and improve predictive performance. The aim is to develop efficient and interpretable methods for understanding complex system dynamics.

Side Projects

I am working with Dr Christopher WoodgateLink opens in a new window on high-entropy alloys (HEAs), focusing on advanced Monte Carlo sampling methods to study their thermodynamic properties and phase behavior. My contributions include implementing efficient sampling algorithms and parallelization strategies in the BraWlLink opens in a new window software package. This work supports the discovery and design of new materials with enhanced physical properties, such as improved stability and resistance to fracture or radiation.

---

---

Contact Details:

E-mail:

hubert.naguszewski@warwick.ac.uk