Personalisation of Computational Models of Attention by Simulated Annealing Parameter Tuning

In this paper it is explored whether personalisation of an existing computational model of attention can increase the model´s validity. Computational models of attention are for instance applied in attention allocation support systems and can benefit from this increased validity. Personalisation is done by tuning the model´s parameters during a training phase, using Simulated Annealing (SA). The adapted attention model is validated using a task, varying in difficulty and attentional demand. Results show that the attention model with personalisation results in a more accurate estimation of an individual´s attention as compared to the model without personalisation.