arXiv subjects

Peter Grindrod

Publications and source records attributed to Peter Grindrod.

12 recordsLinked to original sources

arXiv2025

Word Meanings in Transformer Language Models

We investigate how word meanings are represented in the transformer language models. Specifically, we focus on whether transformer models employ something analogous to a lexical store - where each word has an entry that contains semantic information. To do this, we extracted the token embedding space of RoBERTa-base and k-means clustered it into 200 clusters. In our first study, we then manually inspected the resultant clusters to consider whether they are sensitive to semantic information. In our second study, we tested whether the clusters are sensitive to five psycholinguistic measures: valence, concreteness, iconicity, taboo, and age of acquisition. Overall, our findings were very positive - there is a wide variety of semantic information encoded within the token embedding space. This serves to rule out certain "meaning eliminativist" hypotheses about how transformer LLMs process semantic information.

cs.CL↗

arXiv2024

Modularity, Hierarchical Flows and Symmetry of the Drosophila Connectome

This report investigates the modular organisation of the Central region in the Drosophila connectome. We identify groups of neurones amongst which information circulates rapidly before spreading to the rest of the network using Infomap. We find that information flows along pathways linking distant neurones, forming modules that span across the brain. Remarkably, these modules, derived solely from neuronal connectivity patterns, exhibit a striking left-right symmetry in their spatial distribution as well as in their connections. We also identify a hierarchical structure at the coarse-grained scale of these modules, demonstrating the directional nature of information flow in the system.

q-bio.NC↗

arXiv2024

Dynamical Systems On Generalised Klein Bottles

We propose a high dimensional generalisation of the standard Klein bottle, going beyond those considered previously. We address the problem of generating continuous scalar fields (distributions) and dynamical systems (flows) on such state spaces, which can provide a rich source of examples for future investigations. We consider a class of high dimensional dynamical systems that model distributed information processing within the human cortex, which may be capable of exhibiting some Klein bottle symmetries. We deploy topological data analytic methods in order to analyse their resulting dynamical behaviour, and suggesting future challenges.

math.DS↗

arXiv2023

Diffusion Models for Generative Artificial Intelligence: An Introduction for Applied Mathematicians

Generative artificial intelligence (AI) refers to algorithms that create synthetic but realistic output. Diffusion models currently offer state of the art performance in generative AI for images. They also form a key component in more general tools, including text-to-image generators and large language models. Diffusion models work by adding noise to the available training data and then learning how to reverse the process. The reverse operation may then be applied to new random data in order to produce new outputs. We provide a brief introduction to diffusion models for applied mathematicians and statisticians. Our key aims are (a) to present illustrative computational examples, (b) to give a careful derivation of the underlying mathematical formulas involved, and (c) to draw a connection with partial differential equation (PDE) diffusion models. We provide code for the computational experiments. We hope that this topic will be of interest to advanced undergraduate students and postgraduate students. Portions of the material may also provide useful motivational examples for those who teach courses in stochastic processes, inference, machine learning, PDEs or scientific computing.

cs.LG↗

arXiv2023

Estimating Network Dimension When the Spectrum Struggles

What is the dimension of a network? Here, we view it as the smallest dimension of Euclidean space into which nodes can be embedded so that pairwise distances accurately reflect the connectivity structure. We show that a recently proposed and extremely efficient algorithm for data clouds, based on computing first and second nearest neighbour distances, can be used as the basis of an approach for estimating the dimension of a network with weighted edges. We also show how the algorithm can be extended to unweighted networks when combined with spectral embedding. We illustrate the advantages of this technique over the widely-used approach of characterising dimension by visually searching for a suitable gap in the spectrum of the Laplacian.

cs.SI↗

arXiv2017

A roadmap for the computation of persistent homology

Persistent homology (PH) is a method used in topological data analysis (TDA) to study qualitative features of data that persist across multiple scales. It is robust to perturbations of input data, independent of dimensions and coordinates, and provides a compact representation of the qualitative features of the input. The computation of PH is an open area with numerous important and fascinating challenges. The field of PH computation is evolving rapidly, and new algorithms and software implementations are being updated and released at a rapid pace. The purposes of our article are to (1) introduce theory and computational methods for PH to a broad range of computational scientists and (2) provide benchmarks of state-of-the-art implementations for the computation of PH. We give a friendly introduction to PH, navigate the pipeline for the computation of PH with an eye towards applications, and use a range of synthetic and real-world data sets to evaluate currently available open-source implementations for the computation of PH. Based on our benchmarking, we indicate which algorithms and implementations are best suited to different types of data sets. In an accompanying tutorial, we provide guidelines for the computation of PH. We make publicly available all scripts that we wrote for the tutorial, and we make available the processed version of the data sets used in the benchmarking.

math.AT↗

arXiv2017

A Genetic Algorithm Approach for Modelling Low Voltage Network Demands

Distribution network operators (DNOs) are increasingly concerned about the impact of low carbon technologies on the low voltage (LV) networks. More advanced metering infrastructures provide numerous opportunities for more accurate load flow analysis of the LV networks. However, such data may not be readily available for DNOs and in any case is likely to be expensive. Modelling tools are required which can provide realistic, yet accurate, load profiles as input for a network modelling tool, without needing access to large amounts of monitored customer data. In this paper we outline some simple methods for accurately modelling a large number of unmonitored residential customers at the LV level. We do this by a process we call buddying, which models unmonitored customers by assigning them load profiles from a limited sample of monitored customers who have smart meters. Hence the presented method requires access to only a relatively small amount of domestic customers' data. The method is efficiently optimised using a genetic algorithm to minimise a weighted cost function between matching the substation data and the individual mean daily demands. Hence we can show the effectiveness of substation monitoring in LV network modelling. Using real LV network modelling, we show that our methods perform significantly better than a comparative Monte Carlo approach, and provide a description of the peak demand behaviour.

stat.AP↗

arXiv2016

On Human Consciousness

We consider the implications of the mathematical analysis of neurone-to-neurone dynamical complex networks. We show how the dynamical behaviour of small scale strongly connected networks lead naturally to non-binary information processing and thus multiple hypothesis decision making, even at the very lowest level of the brain's architecture. In turn we build on these ideas to address the hard problem of consciousness. We discuss how a proposed "dual hierarchy model", made up form of both external perceived, physical, elements of increasing complexity, and internal mental elements (experiences), may support a leaning and evolving consciousness. We discuss the idea that a human brain ought to be able to re-conjure subjective mental feelings at will and thus these cannot depend on internal nose (chatter) or internal instability-driven activity. An immediate consequence of this model, grounded in dynamical systems and non-binary information processing, is that finite human brains must always be learning or forgetteing and that any possible subjective internal feeling that may be idealised with a countable infinity of facets, can never be learned by zombies or automata: though it can be experienced more and more fully by an evolving brain (yet never in totality, not even in a lifetime).

q-bio.NC↗

arXiv2014

Opportunities at the Mathematics/Future Cities Interface

We make the case for mathematicians and statisticians to stake their claim in the fast-moving and high-impact research field that is becoming known as Future Cities. After assessing the Future Cities arena, we provide some illustrative challenges where mathematical scientists can make an impact.

cs.CY↗

arXiv2014

Anticipating Activity in Social Media Spikes

We propose a novel mathematical model for the activity of microbloggers during an external, event-driven spike. The model leads to a testable prediction of who would become most active if a spike were to take place. This type of information is of great interest to commercial organisations, governments and charities, as it identifies key players who can be targeted with information in real time when the network is most receptive. The model takes account of the fact that dynamic interactions evolve over an underlying, static network that records who listens to whom. The model is based on the assumption that, in the case where the entire community has become aware of an external news event, a key driver of activity is the motivation to participate by responding to incoming messages. We test the model on a large scale Twitter conversation concerning the appointment of a UK Premier League football club manager. We also present further results for a Bundesliga football match, a marketing event and a television programme. In each case we find that exploiting the underlying connectivity structure improves the prediction of who will be active during a spike. We also show how the half-life of a spike in activity can be quantified in terms of the network size and the typical response rate.

cs.SI↗

arXiv2013

Dynamical Systems to Monitor Complex Networks in Continuous Time

In many settings it is appropriate to treat the evolution of pairwise interactions over continuous time. We show that new Katz-style centrality measures can be derived in this context via solutions to a nonautonomous ODE driven by the network dynamics. This allows us to identify and track, at any resolution, the most influential nodes in terms of broadcasting and receiving information through time dependent links. In addition to the classical notion of attenuation across edges used in the static Katz centrality measure, the ODE also allows for attenuation over time, so that real time "running measures" can be computed. With regard to computational efficiency, we explain why it is cheaper to track good receivers of information than good broadcasters. We illustrate the new measures on a large scale voice call network, where key features are discovered that are not evident from snapshots or aggregates.

cs.SI↗

arXiv2013

Aperiodic dynamics in a deterministic model of attitude formation in social groups

Homophily and social influence are the fundamental mechanisms that drive the evolution of attitudes, beliefs and behaviour within social groups. Homophily relates the similarity between pairs of individuals' attitudinal states to their frequency of interaction, and hence structural tie strength, while social influence causes the convergence of individuals' states during interaction. Building on these basic elements, we propose a new mathematical modelling framework to describe the evolution of attitudes within a group of interacting agents. Specifically, our model describes sub-conscious attitudes that have an activator-inhibitor relationship. We consider a homogeneous population using a deterministic, continuous-time dynamical system. Surprisingly, the combined effects of homophily and social influence do not necessarily lead to group consensus or global monoculture. We observe that sub-group formation and polarisation-like effects may be transient, the long-time dynamics being quasi-periodic with sensitive dependence to initial conditions. This is due to the interplay between the evolving interaction network and Turing instability associated with the attitudinal state dynamics.

physics.soc-ph↗