papers

Publications (20)

cs.CV2025

CrossSDF: 3D Reconstruction of Thin Structures From Cross-Sections

Thomas Walker, Salvatore Esposito, Daniel Rebain +4

Reconstructing complex structures from planar cross-sections is a challenging problem, with wide-reaching applications in medical imaging, manufacturing, and topography. Out-of-the…

cs.CV2026

Beyond Pixel Histories: World Models with Persistent 3D State

Samuel Garcin, Thomas Walker, Steven McDonagh +5

Interactive world models continually generate video by responding to a user's actions, enabling open-ended generation capabilities. However, existing models typically lack a 3D rep…

cs.LG2025

GrokAlign: Geometric Characterisation and Acceleration of Grokking

Thomas Walker, Ahmed Imtiaz Humayun, Randall Balestriero +1

A key challenge for the machine learning community is to understand and accelerate the training dynamics of deep networks that lead to delayed generalisation and emergent robustnes…

hep-ex2024

Terrestrial Very-Long-Baseline Atom Interferometry: Summary of the Second Workshop

Adam Abdalla, Mahiro Abe, Sven Abend +307

This summary of the second Terrestrial Very-Long-Baseline Atom Interferometry (TVLBAI) Workshop provides a comprehensive overview of our meeting held in London in April 2024, build…

quant-ph2020

Improving the Indistinguishability of Single Photons from an Ion-Cavity System

Thomas Walker, Samir Vartabi Kashanian, Travers Ward +1

We investigate schemes for generating indistinguishable single photons, a key feature of quantum networks, from a trapped ion coupled to an optical cavity. Through selection of the…

cs.CV2026

Semantic Foam: Unifying Spatial and Semantic Scene Decomposition

Amr Sharafeldin, Shrisudhan Govindarajan, Thomas Walker +4

Modern scene reconstruction methods, such as 3D Gaussian Splatting, deliver photo-realistic novel view synthesis at real-time speeds, yet their adoption in interactive graphics app…

cs.CV2023

Spherical Feature Pyramid Networks For Semantic Segmentation

Thomas Walker, Varun Anand, Pavlos Andreadis

Semantic segmentation for spherical data is a challenging problem in machine learning since conventional planar approaches require projecting the spherical image to the Euclidean p…

cs.LG2024

Tropical Expressivity of Neural Networks

Paul Lezeau, Thomas Walker, Yueqi Cao +2

We propose an algebraic geometric framework to study the expressivity of linear activation neural networks. A particular quantity of neural networks that has been actively studied…

cs.LG2024

Tightening the Evaluation of PAC Bounds Using Formal Verification Results

Thomas Walker, Alessio Lomuscio

Probably Approximately Correct (PAC) bounds are widely used to derive probabilistic guarantees for the generalisation of machine learning models. They highlight the components of t…

cs.CV2024

Spatially-Adaptive Hash Encodings For Neural Surface Reconstruction

Thomas Walker, Octave Mariotti, Amir Vaxman +1

Positional encodings are a common component of neural scene reconstruction methods, and provide a way to bias the learning of neural fields towards coarser or finer representations…

physics.atom-ph2025

Magnetic Feshbach resonances in Ba+Li collisions due to strong spin-orbit coupling

Masato Morita, Joachim Siemund, Wei Wu +8

We report a pronounced dependence of magnetic Feshbach resonance spectra on the initial hyperfine-Zeeman state of Li in ultracold Ba+Li collisions. The measured num…

cs.LG2026

The Linear Centroids Hypothesis: Features as Directions Learned by Local Experts

Thomas Walker, Ahmed Imtiaz Humayun, Randall Balestriero +1

The Linear Representation Hypothesis (LRH) identifies features of a trained deep network (DN) as linear directions in the activation spaces, i.e., output spaces of intermediate lay…

quant-ph2017

Long-distance single photon transmission from a trapped ion via quantum frequency conversion

Thomas Walker, Koichiro Miyanishi, Rikizo Ikuta +7

Trapped atomic ions are ideal single photon emitters with long lived internal states which can be entangled with emitted photons. Coupling the ion to an optical cavity enables effi…

cs.LG2024

Concept Boundary Vectors

Thomas Walker

Machine learning models are trained with relatively simple objectives, such as next token prediction. However, on deployment, they appear to capture a more fundamental representati…

physics.atom-ph2021

Observation of Feshbach resonances between a single ion and ultracold atoms

Pascal Weckesser, Fabian Thielemann, Dariusz Wiater +6

Controlling physical systems and their dynamics on the level of individual quanta propels both fundamental science and quantum technologies. Trapped atomic and molecular systems, n…

physics.atom-ph2025

Exploring atom-ion Feshbach resonances below the s-wave limit

Fabian Thielemann, Joachim Siemund, Daniel von Schoenfeld +5

Revealing the quantum properties of matter requires a high degree of experimental control accompanied by a profound theoretical understanding. At ultracold temperatures, quantities…

physics.atom-ph2023

Trapping Ion Coulomb Crystals in an Optical Lattice

Daniel Hoenig, Fabian Thielemann, Leon Karpa +3

We report the optical trapping of multiple ions localized at individual lattice sites of a one-dimensional optical lattice. We observe a fivefold increase in robustness against axi…

cs.CV2023

Explicit Neural Surfaces: Learning Continuous Geometry With Deformation Fields

Thomas Walker, Octave Mariotti, Amir Vaxman +1

We introduce Explicit Neural Surfaces (ENS), an efficient smooth surface representation that directly encodes topology with a deformation field from a known base domain. We apply t…

cond-mat.quant-gas2026

A High-Flux Source of Cold Strontium with a Loading Rate of atoms/s for Open Release

Thomas Walker, Anna L. Marchant, Elliot Bentine +17

We present a high-flux source of cold strontium atoms based on a two-dimensional magneto-optical trap (2D MOT) and a Zeeman slower. We use the source to load a 3D MOT in a separate…

cs.LG2026

The Geometric Structure of Models Learning Sparse Data

Thomas Walker, T. Mitchell Roddenberry, Ahmed Imtiaz Humayun +2

The manifold hypothesis (MH) is often used to explain how machine learning can overcome the curse of dimensionality. However, the MH is only applicable in regimes where the trainin…