arXiv subjects

Blaise Delaney

Publications and source records attributed to Blaise Delaney.

5 recordsLinked to original sources

arXiv2026

Capturing Cardiac Cyclicity through Phase-Equivariant Self-Supervised Learning

The cyclic structure of physiological processes offers a natural prior for self-supervised representation learning, and the cardiac cycle provides a particularly well-defined setting in which to exploit it. We derive a phase-equivariant self-supervised objective and introduce Winder, a joint-embedding architecture that organises representations into phase-invariant coordinates and phase-rotating harmonic subspaces. Its transport operator is fixed and closed-form, derived from the cycle's geometry rather than learned, and adds no parameters. Evaluated on PTB-XL under a frozen linear-probe protocol, Winder attains diagnostic accuracy within the range reported by state-of-the-art self-supervised methods at a ~1 M parameter footprint, while exhibiting phase-equivariant latent geometry. These findings demonstrate that explicitly encoding cardiac-phase symmetry can preserve diagnostically useful information while yielding a latent geometry that is legible, parameter-efficient, and directly tied to a measurable physiological quantity.

cs.LG↗

arXiv2026

Decorrelation of neural networks from particle lifetimes in the LHCb topological $b$ trigger

The LHCb topological beauty trigger is the primary set of algorithms for selecting collision events containing $b$-hadrons in the fully software-based LHCb trigger. The algorithms apply monotonic Lipschitz neural networks (NNs) to select vertices of charged particles consistent with the distinct topology of a $b$ decay, i.e., those with large lifetimes and transverse momentum. Many analyses of the events recorded require that the selection must be unbiased with respect to the $b$-hadron lifetime at large lifetimes. Accurate reconstruction is challenging in busier detector environments, in which several visible proton-proton collisions occur simultaneously per bunch crossing, such that misassociation of decay products can result in vertices with artificially large measured lifetimes. This paper presents two approaches to mitigate correlations between NN scores and candidate lifetimes at large lifetime, and evaluates the performance of the resulting models.

hep-ex↗

arXiv2026

Sonata: A Hybrid World Model for Inertial Kinematics under Clinical Data Scarcity

We introduce Sonata, a compact latent world model for six-axis trunk IMU representation learning under clinical data scarcity. Clinical cohorts typically comprise tens to hundreds of patients, making web-scale masked-reconstruction objectives poorly matched to the problem. Sonata is a 3.77 M-parameter hybrid model, pre-trained on a harmonised corpus of nine public datasets (739 subjects, 190k windows) with a latent world-model objective that predicts future state rather than reconstructing raw sensor traces. In a controlled comparison against a matched autoregressive forecasting baseline (MAE) on the same backbone, Sonata yields consistently stronger frozen-probe clinical discrimination, prospective fall-risk prediction, and cross-cohort transfer across a 14-arm evaluation suite, while producing higher-rank, more structured latent representations. At 3.77 M parameters the model is compatible with on-device wearable inference, offering a step toward general kinematic world models for neurological assessment.

cs.LG↗

arXiv2025

Seeing the Forest Through the Trees: Knowledge Retrieval for Streamlining Particle Physics Analysis

Generative Large Language Models (LLMs) are a promising approach to structuring knowledge contained within the corpora of research literature produced by large-scale and long-running scientific collaborations. Within experimental particle physics, such structured knowledge bases could expedite methodological and editorial review. Complementarily, within the broader scientific community, generative LLM systems grounded in published work could make for reliable companions allowing non-experts to analyze open-access data. Techniques such as Retrieval Augmented Generation (RAG) rely on semantically matching localized text chunks, but struggle to maintain coherent context when relevant information spans multiple segments, leading to a fragmented representation devoid of global cross-document information. Here, we utilize the hierarchical organization of experimental physics articles to build a tree representation of the corpus, and present the SciTreeRAG system that uses this structure to create contexts that are more focused and contextually rich than standard RAG. Additionally, we develop methods for using LLMs to transform the unstructured corpus into a structured knowledge graph representation. We then implement SciGraphRAG, a retrieval system that leverages this knowledge graph to access global cross-document relationships eluding standard RAG, thereby encapsulating domain-specific connections and expertise. We demonstrate proof-of-concept implementations using the corpus of the LHCb experiment at CERN.

hep-ex↗

arXiv2023

Applications of Lipschitz neural networks to the Run 3 LHCb trigger system

The operating conditions defining the current data taking campaign at the Large Hadron Collider, known as Run 3, present unparalleled challenges for the real-time data acquisition workflow of the LHCb experiment at CERN. To address the anticipated surge in luminosity and consequent event rate, the LHCb experiment is transitioning to a fully software-based trigger system. This evolution necessitated innovations in hardware configurations, software paradigms, and algorithmic design. A significant advancement is the integration of monotonic Lipschitz neural networks into the LHCb trigger system. These deep learning models offer certified robustness against detector instabilities, and the ability to encode domain-specific inductive biases. Such properties are crucial for the inclusive heavy-flavour triggers and, most notably, for the topological triggers designed to inclusively select $b$-hadron candidates by exploiting the unique kinematic and decay topologies of beauty decays. This paper describes the recent progress in integrating Lipschitz neural networks into the topological triggers, highlighting the resulting enhanced sensitivity to highly displaced multi-body candidates produced within the LHCb acceptance.

hep-ex↗