Artificial Intelligence Institute

Deep learning

Deep learning is a specific type of machine learning whereby artificial neural networks learn from complex data such as images, video, and audio. Neural networks can be very powerful but are energy intensive due to the computational requirements.

We are currently focused on three particular problems in deep learning research and applications: user-friendly deep learning, explainability of deep learning, and accurate quantification of predictive uncertainty. We are also working on continual deep learning, which is closely related to our work on real-time analytics.

Real-time analytics

Real-time analytics for big data provides users with timely insights so they can respond without delay. It requires a step-change in computational performance but has huge potential to both transform existing businesses and create opportunities for new businesses to evolve. Continual learning is the next step, where learning is retained.

Open source software

A key focus of the Artificial Intelligence Institute is developing and maintaining open source software to support research, innovation and the application of AI through fast, reliable and secure development tools.

Open source software reduces training costs and provides access to publicly available frameworks, workflows, data sets and models, alongside a community providing monitoring and security.