Nian Ran, a prospective PhD student at AMBS, shares his thoughts from a recent major global computation conference.
In July 2024, the Genetic and Evolutionary Computation Conference (GECCO) held its annual conference in Melbourne, Australia, and was attended by a number of AMBS academics. The event brings together researchers, practitioners, and enthusiasts from around the globe to explore the latest advancements in genetic and evolutionary computation.
About GECCO
GECCO is a critical platform for exchanging ideas, fostering collaborations, and showcasing cutting-edge research in the fields of evolutionary algorithms, genetic programming, and related areas.
It features a diverse range of topics such as optimisation techniques, machine learning applications, bioinformatics, robotics, and artificial life - all areas which are at the very forefront of the technological revolution spreading across the world.
One of the unique aspects of GECCO is its inclusive approach, which encourages participation from a broad spectrum of disciplines.
Insights from the conference
Attendees gain insights into how evolutionary computation can solve complex real-world problems. The conference included special sessions that highlighted practical implementations and innovative solutions developed using evolutionary techniques.
Examples of real-world problems include optimising logistics networks, improving healthcare data analysis, and enhancing molecular design. This work is so important because it provides efficient and effective solutions to critical challenges where limited resources, data, and evaluation budgets are available.
Presenting a paper at GECCO
At this year’s conference I also presented a paper, 'Multi-objective evolutionary GAN for tabular data synthesis' co-authored with colleagues Richard Allmendinger, Bahrul Ilmi Nasution, Claire Little and Mark Elliot. The paper addresses a significant challenge in the field of data science - namely how to generate high-quality synthetic tabular data that balances utility and privacy.
The motivation behind our work stems from the growing need for synthetic data in various applications, such as data sharing and privacy-preserving data analysis between organisations.
Utility refers to how effectively synthetic tabular data can be used for training machine learning models, while privacy indicates the risk of leaking real information from the original data. Our framework can be used to generate synthetic tabular data that maintains training effectiveness while preventing data leaks, and provides practical solutions for enhancing data privacy without compromising on the quality of synthetic data.
Future work
We now plan to build on this work by exploring the use of diffusion models in tabular data synthesis. While diffusion models have demonstrated outstanding performance in generating images, their application to tabular data remains largely unexplored.
This work is relevant to everyday business because it can help companies share data safely and effectively. We hope our work will facilitate more effective and lower-risk research data exchange between different organisations.
Reflections
Presenting our paper at the GECCO conference was incredibly exciting - not only because it is one of the premier conferences in evolutionary computation, but also because of its vibrant and passionate community and committee.
In particular, the conference provides ample opportunities to connect with researchers from various backgrounds through a wide range of events designed to foster networking and collaboration.
The conference’s community-oriented atmosphere also makes it an ideal venue for both seasoned researchers and newcomers to the field to connect, learn, and contribute to the ever-evolving landscape of genetic and evolutionary computation.
You can read the paper 'Multi-objective evolutionary GAN for tabular data synthesis' on the arXiv website.
