Professor Plamen Angelov of Lancaster’s School of Computing and Communications recently delivered a keynote presentation on real-time visual classification at the IEEE World Congress on Computational Intelligence (IEEE WCCI) in Yokohama, Japan. The IEEE WCCI has become the world’s largest event in computational intelligence since its inception 30 years ago, and welcomes experts in AI and machine learning from across the globe. The event is sponsored and hosted by the IEEE itself, a technical and professional organisation that has been in operation for 140 years and strives to advance technology for the betterment of humanity.

Professor Angelov’s paper centred on the union of machine learning and AI methods for logic and reasoning. “Deep Learning” (a form of complex machine learning that utilises neural networks to emulate human decision-making) has become the focus of contemporary Machine Learning, attracting the attention and interest not only of the wider scientific community and industry, but also society and policymakers fuelled by the remarkable achievements of foundational models such as large language models (LLMs). However, Professor Angelov also argues that more standard AI tools are overlooked, despite their use in addressing some of the issues associated with Deep Learning. He proposes the alternative, prototype-based deep learning models which he and other researchers at Lancaster have pioneered, as a potential solution.

Prototype-based deep learning models work by comparing inputted source data with external data, and analysing the similarity between the source and the external data in order to make decisions, effectively blending newer Deep Learning models with more traditional logic-based modes of decision-making. Professor Angelov and his team at Lancaster have already successfully employed a prototype-based deep learning model in the detection of Deepfaked images. This process is a lot more transparent than what is used by many other types of Deep Learning (where it can often be difficult to understand the decision-making process behind its conclusions), and better emulates human decision-making than its predecessors.

The event itself was attended by over 3000 people. On having the opportunity to present his paper, Professor Angelov said: “This was a great opportunity to present to a large audience including top researchers in the field, and bring the latest research results and vision for the future of this fast-developing area. WCCI was an excellent forum for my talk since it brings together the two somewhat different wings and approaches towards AI, namely the (currently very popular) Machine Learning approach which is a more statistics-based one and the older, more traditional and, I would say, somewhat overlooked at the moment, symbolic, logic and reasoning based one. My view is that, similarly to the way the human brain works, both approaches can symbiotically co-exist. They each have their own advantages and limits and, in my view, the future lies in bringing these closer, which would ideally result in a more human-like and human-oriented machine learning and AI methods, algorithms and, eventually, products. I was also delighted to meet many of my colleagues with whom we have been researching in this area for over 35 years.”