Rational machines that can learn throughout their life
Prof Alessandra Russo works in machine learning, looking specifically at algorithms that enable machines to learn and adapt from their environment continuously. These algorithms would allow machines to have a natural-language dialogue with a human user about what has been learnt and why decisions have been made. Crucially, this requires the machine to process information provided back to it from the human user about any errors that it’s made, and turn these into data which it can use to continue learning. The learning process is, therefore, one of continuous improvement, focusing on the areas in which the improvement is needed.
Alessandra’s algorithms mean that new data – about problem areas – is continuously being fed back to the machine. This means that the types of things that we know via “common sense” – but that machines find very difficult to process – can eventually be learnt. This is different to most machine learning algorithms because these are trained on a (sometimes very large) set of data.
In the future, these machine-learning algorithms can be applied to a number of industries, including manufacturing, retail, energy networks or computing.
Other promising applications include:
- Assistive technologies for the blind, to help for example navigating in an unfamiliar and changing environment.