AI for Science, Engineering & Commerce: the intelligent selection of experiments, models and methods
Acquiring new data and shedding new light on old data both require researchers to make selection of experiments, models and methods. All forms of data recording and computation have costs associated with them: economic costs, hardware and memory limitations, and time. This theme aims to automate the process of choosing which tools to use while at the same time taking the costs into account, enabling efficient, optimal experimentation to be performed with a given budget. Methods used include Continual Reinforcement Learning, models which continually explore and re-develop better strategies, prevent catastrophic forgetting from the past and improve data efficiency when turning to new data.
Please contact the Principal Investigator for further information: