arXiv subjects

Andrew McDowell

Publications and source records attributed to Andrew McDowell.

8 recordsLinked to original sources

arXiv2025

AI as a teaching tool and learning partner

The arrival of AI tools and in particular Large Language Models (LLMs) has had a transformative impact on teaching and learning and institutes are still trying to determine how to integrate LLMs into education in constructive ways. Here, we explore the adoption of LLM-based tools into two teaching programmes, one undergraduate and one postgraduate. We provided to our classes (1) a LLM-powered chatbot that had access to course materials by RAG and (2) AI-generated audio-only podcasts for each week$\text{'}$s teaching material. At the end of the semester, we surveyed the classes to gauge attitudes towards these tools. The classes were small and from biological courses. The students felt positive about AI generally and that AI tools made a positive impact on teaching. Students found the LLM-powered chatbot easy and enjoyable to use and felt that it enhanced their learning. The podcasts were less popular and only a small proportion of the class listened weekly. The class as a whole was indifferent to whether the podcasts should be used more widely across courses, but those who listened enjoyed them and were in favour.

cs.HC↗

arXiv2024

Classification of Spontaneous and Scripted Speech for Multilingual Audio

Distinguishing scripted from spontaneous speech is an essential tool for better understanding how speech styles influence speech processing research. It can also improve recommendation systems and discovery experiences for media users through better segmentation of large recorded speech catalogues. This paper addresses the challenge of building a classifier that generalises well across different formats and languages. We systematically evaluate models ranging from traditional, handcrafted acoustic and prosodic features to advanced audio transformers, utilising a large, multilingual proprietary podcast dataset for training and validation. We break down the performance of each model across 11 language groups to evaluate cross-lingual biases. Our experimental analysis extends to publicly available datasets to assess the models' generalisability to non-podcast domains. Our results indicate that transformer-based models consistently outperform traditional feature-based techniques, achieving state-of-the-art performance in distinguishing between scripted and spontaneous speech across various languages.

cs.CL↗

arXiv2018

Contagious sets in a degree-proportional bootstrap percolation process

We study the following bootstrap percolation process: given a connected graph $G$, a constant $ρ\in [0, 1]$ and an initial set $A \subseteq V(G)$ of \emph{infected} vertices, at each step a vertex~$v$ becomes infected if at least a $ρ$-proportion of its neighbours are already infected (once infected, a vertex remains infected forever). Our focus is on the size $h_ρ(G)$ of a smallest initial set which is \emph{contagious}, meaning that this process results in the infection of every vertex of $G$. Our main result states that every connected graph $G$ on $n$ vertices has $h_ρ(G) < 2ρn$ or $h_ρ(G) = 1$ (note that allowing the latter possibility is necessary because of the case $ρ\leq\tfrac{1}{2n}$, as every contagious set has size at least one). This is the best-possible bound of this form, and improves on previous results of Chang and Lyuu and of Gentner and Rautenbach. We also provide a stronger bound for graphs of girth at least five and sufficiently small $ρ$, which is asymptotically best-possible.

math.CO↗

arXiv2018

Hamilton $\ell$-cycles in randomly-perturbed hypergraphs

We prove that for integers $2 \leq \ell < k$ and a small constant $c$, if a $k$-uniform hypergraph with linear minimum codegree is randomly `perturbed' by changing non-edges to edges independently at random with probability $p \geq O(n^{-(k-\ell)-c})$, then with high probability the resulting $k$-uniform hypergraph contains a Hamilton $\ell$-cycle. This complements a recent analogous result for Hamilton $1$-cycles due to Krivelevich, Kwan and Sudakov, and a comparable theorem in the graph case due to Bohman, Frieze and Martin.

math.CO↗

arXiv2017

Dispersion processes

We study a synchronous dispersion process in which $M$ particles are initially placed at a distinguished origin vertex of a graph $G$. At each time step, at each vertex $v$ occupied by more than one particle at the beginning of this step, each of these particles moves to a neighbour of $v$ chosen independently and uniformly at random. The dispersion process ends once the particles have all stopped moving, i.e. at the first step at which each vertex is occupied by at most one particle. For the complete graph $K_n$ and star graph $S_n$, we show that for any constant $δ>1$, with high probability, if $M \le n/2(1-δ)$, then the process finishes in $O(\log n)$ steps, whereas if $M \ge n/2(1+δ)$, then the process needs $e^{Ω(n)}$ steps to complete (if ever). We also show that an analogous lazy variant of the process exhibits the same behaviour but for higher thresholds, allowing faster dispersion of more particles. For paths, trees, grids, hypercubes and Cayley graphs of large enough sizes (in terms of $M$) we give bounds on the time to finish and the maximum distance traveled from the origin as a function of the number of particles $M$.

cs.DM↗

arXiv2016

Triangle-tilings in graphs without large independent sets

We study the minimum degree necessary to guarantee the existence of perfect and almost-perfect triangle-tilings in an $n$-vertex graph $G$ with sublinear independence number. In this setting, we show that if $δ(G) \ge n/3 + o(n)$ then $G$ has a triangle-tiling covering all but at most four vertices. Also, for every $r \ge 5$, we asymptotically determine the minimum degree threshold for a perfect triangle-tiling under the additional assumptions that $G$ is $K_r$-free and $n$ is divisible by $3$.

math.CO↗

arXiv2014

DNA origami and the complexity of Eulerian circuits with turning costs

Building a structure using self-assembly of DNA molecules by origami folding requires finding a route for the scaffolding strand through the desired structure. When the target structure is a 1-complex (or the geometric realization of a graph), an optimal route corresponds to an Eulerian circuit through the graph with minimum turning cost. By showing that it leads to a solution to the 3-SAT problem, we prove that the general problem of finding an optimal route for a scaffolding strand for such structures is NP-hard. We then show that the problem may readily be transformed into a Traveling Salesman Problem (TSP), so that machinery that has been developed for the TSP may be applied to find optimal routes for the scaffolding strand in a DNA origami self-assembly process. We give results for a few special cases, showing for example that the problem remains intractable for graphs with maximum degree 8, but is polynomial time for 4-regular plane graphs if the circuit is restricted to following faces. We conclude with some implications of these results for related problems, such as biomolecular computing and mill routing problems.

math.CO↗

arXiv2013

Non-Vertex-Balanced Factors in Random Graphs

We prove part of a conjecture by Johansson, Kahn and Vu \cite{JKV} regarding threshold functions for the existence of an $H$-factor in a random graph \gnp. We prove that the conjectured threshold function is correct for any graph $H$ which is not covered by its densest subgraphs. We also demonstrate that the main result of \cite{JKV} generalises to multigraphs, digraphs, and a multipartite model.

math.CO↗