arXiv subjects

Eric J. Hall

Publications and source records attributed to Eric J. Hall.

4 recordsLinked to original sources

arXiv2024

Using data collected from structured light plethysmography to differentiate breathing pattern disorder from normal breathing: A study group report

This report relates to a study group hosted by the EPSRC funded network, Integrating data-driven BIOphysical models into REspiratory MEdicine (BIOREME), and supported by SofTMech and Innovate UK, Business Connect. This report summarises the work undertaken on a challenge presented by two of the authors, Mathew Bulpett and Dr Emily Fraser. The aim was to identify approaches to analyse data collected using structured light plethysmography (SLP) from (n=31) healthy volunteers and (n=67) patients with Breathing Pattern Disorder (BPD) attributed to "long COVID", i.e. post-acute COVID-19 sequelae. This report explores several approaches including dimensionality reduction techniques on the available data and alternative indices extracted from variation in the time-series data for each measurement. Further proposals are also outlined such as different spatial indices that could be extracted from the SLP data, and the potential to couple to mechanical models of the lungs, chest and abdomen. However, running these latter analyses was beyond the scope of the limited study group timeframe. This exploratory analysis did not identify any clear SLP biomarkers of BPD in these cohorts, however recommendations are made for using SLP technologies in future BPD studies based on its findings.

physics.med-ph↗

arXiv2021

Mutual Information for Explainable Deep Learning of Multiscale Systems

Timely completion of design cycles for complex systems ranging from consumer electronics to hypersonic vehicles relies on rapid simulation-based prototyping. The latter typically involves high-dimensional spaces of possibly correlated control variables (CVs) and quantities of interest (QoIs) with non-Gaussian and possibly multimodal distributions. We develop a model-agnostic, moment-independent global sensitivity analysis (GSA) that relies on differential mutual information to rank the effects of CVs on QoIs. The data requirements of this information-theoretic approach to GSA are met by replacing computationally intensive components of the physics-based model with a deep neural network surrogate. Subsequently, the GSA is used to explain the network predictions, and the surrogate is deployed to close design loops. Viewed as an uncertainty quantification method for interrogating the surrogate, this framework is compatible with a wide variety of black-box models. We demonstrate that the surrogate-driven mutual information GSA provides useful and distinguishable rankings on two applications of interest in energy storage. Consequently, our information-theoretic GSA provides an "outer loop" for accelerated product design by identifying the most and least sensitive input directions and performing subsequent optimization over appropriately reduced parameter subspaces.

cs.LG↗

arXiv2020

GINNs: Graph-Informed Neural Networks for Multiscale Physics

We introduce the concept of a Graph-Informed Neural Network (GINN), a hybrid approach combining deep learning with probabilistic graphical models (PGMs) that acts as a surrogate for physics-based representations of multiscale and multiphysics systems. GINNs address the twin challenges of removing intrinsic computational bottlenecks in physics-based models and generating large data sets for estimating probability distributions of quantities of interest (QoIs) with a high degree of confidence. Both the selection of the complex physics learned by the NN and its supervised learning/prediction are informed by the PGM, which includes the formulation of structured priors for tunable control variables (CVs) to account for their mutual correlations and ensure physically sound CV and QoI distributions. GINNs accelerate the prediction of QoIs essential for simulation-based decision-making where generating sufficient sample data using physics-based models alone is often prohibitively expensive. Using a real-world application grounded in supercapacitor-based energy storage, we describe the construction of GINNs from a Bayesian network-embedded homogenized model for supercapacitor dynamics, and demonstrate their ability to produce kernel density estimates of relevant non-Gaussian, skewed QoIs with tight confidence intervals.

physics.comp-ph↗

arXiv2011

Partial choice functions for families of finite sets

Let m>2 be an integer. We show that ZF + "For every integer n, Every countable family of non-empty sets of cardinality at most n has an infinite partial choice function" is not strong enough to prove that every countable set of m-element sets has a choice function. In the case where m=p is prime, to obtain the independence result we make use of a permutation model in which the set of atoms has the structure of a vector space over the field of p elements. When m is non-prime, a suitable permutation model is built from the models used in the prime cases.

math.LO↗