The Artificial Intelligence, Machine Learning, and TinyML Group

Group leader: Prof Absalom Ezugwu

Prof Ezugwe is interested in artificial intelligence, global optimization, machine learning, metaheuristic algorithm design, deep learning, medical image analysis, tinyml, data mining, and clustering. With over 15 years of academic experience, he has made significant contributions to these fields.

Absalom is currently mentoring several postgraduate students and seeking potential PhD candidates interested in conducting cutting-edge research in the field of artificial intelligence.

Key research areas:

• Developing assistive technologies for people with disabilities, such as real-time sign language translation.
• Soil moisture and crop health monitoring using TinyML-enabled sensors.
• Deploying TinyML models on low-power cameras and sensors to monitor animal behaviour and biodiversity.
• Energy-efficient AI for smart agriculture and climate change mitigation.
• Real-time computer vision (e.g., object detection, gesture recognition)

Automated Neural Networks and Pattern Recognition Group

Group leader: Prof Tiny du Toit

Prof du Toit is interested in constructing feedforward and recurrent neural networks that is often a trial-and-error activity as it requires the specification of many hyperparameters, which can be difficult and time-consuming for some data scientists and practitioners. He aims to develop novel end-to-end frameworks to automate the construction of these complex neural network architectures and make the techniques accessible to more data scientists and practitioners. In addition, he applies the neural network models to different novel problem areas, which can benefit from the neural networks' highly accurate pattern recognition capabilities.

Key research areas:

• Affective computing and deep learning to perform sentiment analysis
• Classifying spam with generalised additive neural networks
• Detecting Network-based IoT Botnet Attacks using Supervised Autoencoders
• Radial Basis Function Neural Networks Towards Electronic Trust Quantification
• Automated Terrain Classification with a Bayesian Hyperparameter Optimized Deep Supervised Autoencoder Model