Glass optimization for optical design: CMA-ES optimizer with integer handling

Traditional optimization of lens material uses a two dimensional continuous space to describe a material along with a local optimizer like Levenberg-Marquardt. However other algorithms may be more suited for harder problems, in particular if the problem has several local minima, large number of variables and integer variables. The presented optimization method make use of evolution strategy with covariance matrix adaptation (CMA-ES). A modified version of this algorithm is used to handle the optimization of lens material as integer variables. Results will be focused on the performance obtained with complex camera lens composed of 31 diopters and optimized for 3 configurations corresponding to 3 different f number