|Title:||Fitness approximation in expensive optimization problems|
Although evolutionary algorithms have been proved powerful for complex optimization problems, their applications in industrial scenarios may suffer from costly evaluations of trial solutions, which are often implemented through a time-consuming procedure such as system simulation. This thesis will aim to develop an enhanced evolutionary algorithm with embedded fitness approximation to reduce computational expenses on function evaluations. On one hand, a fitness approximate model will be built incrementally based on the results obtained from simulations as original evaluations of solutions. At the same time the fitness model will be utilized to assess some individuals in the population during the running of the algorithm. A crucial strategy will be investigated on when and where to use the fitness model in replacement of simulations to increase the time efficiency while still acquiring high quality of solutions.
The thesis project will be jointly supervised by Q-TAGG and MDH.
A student needs to have solid knowledge in math, machine learning and optimization as well as good programming skills for conducting this project. Q-TAGG will follow a procedure of qualification testing to select a qualified candidate.
Please contact George Fodor (firstname.lastname@example.org) if you are interested
|Prerequisites:||Solid knowledge in math, machine learning and optimization, good programming skills|
|IDT supervisors:||Ning Xiong|
|Company contact:||Q-TAGG George Fodor email@example.com|