Christopher Wesley Cleghorn

Christopher Wesley Cleghorn

Auto-generated profile, based on publicly available information

Christopher Wesley Cleghorn

University of the Witwatersrand, Johannesburg, South Africa

Research Focus & Metrics

Academic Career Length

13Years (2014-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
27 September 2017 in Swarm Intelligence
Verified
Journal Article
Closed access
PDF file
CCChristopher W. Cleghorn
Christopher W. Cleghorn
AEAndries P. Engelbrecht
Andries P. Engelbrecht

14 July 2024
Verified
Conference Paper
Open access
MNMuhammad Umair Nasir
Muhammad Umair Nasir
SESam Earle
Sam Earle
JTJulian Togelius
Julian Togelius
SJSteven James
Steven James
CCChristopher Cleghorn
Christopher Cleghorn
Large language models (LLMs) have emerged as powerful tools capable of accomplishing a broad spectrum of tasks. Their abilities span numerous areas, and one area where they have made a significant impact is in the domain of code generation. Here, we propose using the coding abilities of LLMs to introduce meaningful variations to code defining neural networks. Meanwhile, Quality-Diversity (QD) algorithms are known to discover diverse and robust solutions. By merging the code-generating abilities of LLMs with the diversity and robustness of QD solutions, we introduce LLMatic, a Neural Architecture Search (NAS) algorithm. While LLMs struggle to conduct NAS directly through prompts, LLMatic uses a procedural approach, leveraging QD for prompts and network architecture to create diverse and high-performing networks. We test LLMatic on the CIFAR-10 and NAS-bench-201 benchmarks, demonstrating that it can produce competitive networks while evaluating just 2, 000 candidates, even without prior knowledge of the benchmark domain or exposure to any previous top-performing models for the benchmark. The open-sourced code is available at https://github.com/umair-nasir14/LLMatic.

14 January 2014 in Swarm Intelligence
Verified
Journal Article
Closed access
PDF file
CCChristopher W. Cleghorn
Christopher W. Cleghorn
Andries EngelbrechtAndries Engelbrecht
Andries Engelbrecht

2 June 2015 in Swarm Intelligence
Verified
Journal Article
Closed access
CCChristopher W. Cleghorn
Christopher W. Cleghorn
Andries EngelbrechtAndries Engelbrecht
Andries Engelbrecht

7 October 2015 in Swarm Intelligence
Verified
Journal Article
Open access
PDF file
BLBarend J. Leonard
Barend J. Leonard
Andries EngelbrechtAndries Engelbrecht
Andries Engelbrecht
CCChristopher W. Cleghorn
Christopher W. Cleghorn
This article investigates various aspects of angle modulated particle swarm optimisers (AMPSO). Previous attempts at improving the algorithm have only been able to produce better results in a handful of test cases. With no clear understanding of when and why the algorithm fails, improving the algorithm’s performance has proved to be a difficult and sometimes blind undertaking. Therefore, the aim of this study is to identify the circumstances under which the algorithm might fail, and to understand and provide evidence for such cases. It is shown that the general assumption that good solutions are grouped together in the search space does not hold for the standard AMPSO algorithm or any of its existing variants. The problem is explained by specific characteristics of the generating function used in AMPSO. Furthermore, it is shown that the generating function also prevents particle velocities from decreasing, hindering the algorithm’s ability to exploit the binary solution space. Methods are proposed to both confirm and potentially solve the problems found in this study. In particular, this study addresses the problem of finding suitable generating functions for the first time. It is shown that the potential of a generating function to solve arbitrary binary optimisation problems can be quantified. It is further shown that a novel generating function with a single coefficient is able to generate solutions to binary optimisation problems with fewer than four dimensions. The use of ensemble generating functions is proposed as a method to solve binary optimisation problems with more than 16 dimensions.

19 August 2019 in Swarm Intelligence
Verified
Journal Article
Closed access
PDF file
CSChristiaan Scheepers
Christiaan Scheepers
Andries EngelbrechtAndries Engelbrecht
Andries Engelbrecht
CCChristopher W. Cleghorn
Christopher W. Cleghorn