There is a growing trend towards personalisation of services in many real-world application domains. This involves learning user preferences over alternative choices, e.g., products, movies and hotels. Ridesharing, for example,  is used as a remedy to reduce travel costs, traffic congestion and pollution. Mojtaba has proposed a method to learn user preferences to recommend better matching between drivers and passengers.

There are some challenges in preference learning regarding the scaling. Motjaba and the team are analysing more robust forms of preference inferences that are invariant to the scaling of features, inputs, or both simultaneously.

This work has other potential applications in decision-making such as disease diagnosis.