MOO publication using CMA-ES variant using rank-based Multi-Criteria Decision Making

[Nicolas Robidoux]

You may find the CVPR 2023 main paper and supplement found at https://light.princeton.edu/publication/lidar-in-the-loop-hyperparameter-optimization/ of interest, in particular Sections 4.2 and 5.1 of the main paper, and  Sections 1.1-1.6 and 3.2-3.3 the Supplement.

CMA-ES variant was implemented in private branch of DEAP. (Thank you, DEAP people!)

[Nicolas Robidoux]

The Mosleh et al. CMA-ES method referred to in the paper, also built into a private version of DEAP, as well as the Robidoux et al. variant, are discussed here: https://light.princeton.edu/publication/hil_image_optimization/ https://light.princeton.edu/publication/hdr_isp_opt/

We have been very successful using "very active" weights, that is, centroid weights which a significant proportion of negative weights. See Section 3.3 of the Supplement of the LiDAR optimization paper https://light.princeton.edu/publication/lidar-in-the-loop-hyperparameter-optimization/