An investigation of genetic programming for Q-function approximation
Đào Ngọc Phong, Nguyễn Xuân Hoài, Nguyễn Thanh Thủy, Nguyen Quang Uy · Journal of Computer Science and Cybernetics · 2012
In this paper, we investigate the use of two different fitness functions, MAE and RAE, for tree adjoining grammar guided genetic programming (TAG3P) in solving the problem of Gaussian Q-function approximation. The results show that these different fitness functions have different effects on the quality of approximations obtained by TAG3P. Moreover, the results also show that approximations found by TAG3P are better than all but one human expert designed approximations in the literature. This encourages further studies into the application of TAG3P (and GP) in solving the problem of Q-function approximation.