Samineh Bagheri is our Leiden Institute of Advanced Computer Science’s nominee for the C.J. Kok Jury Award 2020.
About Samineh's research
Say you want to model how an aeroplane wing functions. Or how a car behaves when subjected to a crash test. Here, regular mathematical equations do not cut the mustard. You need complex simulations called ‘black boxes’ to get the desired result. The term black box refers to the lack of explicit algebraic formulations within these simulations. Because they have to take into account a lot of parameters, running black-box simulations requires a lot of time and computing power. In the case of modelling an aeroplane wing you can think of parameters like size, shape and the material of which the wing is made. Moreover, finding an optimal design requires not just one simulation, but many trials. This makes it even more time-consuming.
‘Life is similar to complex black-box optimisation processes with phases of exploration, exploitation and decision making under uncertainty, with conflicting objectives and constraints. Exploring the unknown uncomfortable paths will help us to have a better perception of life’s landscape.’
In her PhD research, Samineh developed ways to reduce the time and computing power required. Not by changing the simulations, but by finding an optimal design with as few simulations as possible. Samineh for instance came up with an algorithm that uses previous designs to develop the next. Such step by step improvements prevent you from having to randomly try different parameters until you find the optimal aeroplane wing, or car design. That would take forever.