Azarbaijan Shahid Madani University Communications in Combinatorics and Optimization 2538-2128 Articles in Press 2025 09 10 A new quasi-Newton algorithm for constructing the Pareto front of multiobjective optimization problems by implementing warm-start strategies 15014 10.22049/cco.2025.30421.2459 EN Fereshteh Akbari Faculty of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran Esmaile Khorram Faculty of Mathematics and Computer Science, Amirkabir University of Technology, Tehran, Iran Mehrdad Ghaznavi Faculty of Mathematical Sciences, Shahrood University of Technology, Shahrood, Iran Journal Article 2025 03 14 Many numerical procedures for finding efficient solutions of multiobjective optimization problems are variants of Newton method, that utilize the Hessian matrix of second derivatives. Quasi-Newton methods are used for situations in which the calculation of the Hessian matrix or its inverse is difficult or expensive. In the quasi-Newton methods, only first derivatives are utilized to build an approximation of the actual Hessian matrix over a number of iterations. One of the weaknesses of Newton and quasi-Newton methods is choosing the proper starting points. In fact, the starting points should be close enough to the nondominated solution to have at least quadratic convergence. Therefore, in this study, by applying the convex hull of the individual minimums (CHIMs), we present a procedure for selecting an appropriate starting point for the quasi-Newton method with the BFGS (Broyden, Fletcher, Goldfarb and Shanno) approximation. Moreover, a new algorithm for constructing a uniform approximation of the Pareto front is presented, which can produce more than one efficient point located on the Pareto front in each iteration. To comprehensively compare the proposed algorithm with existing algorithms, three indices are considered: purity metric, measures of coverage, and spacing metric. Extensive numerical experiments show the significant advantage of the proposed algorithm. Moreover, the obtained boundary approximation follows an almost uniform distribution. https://comb-opt.azaruniv.ac.ir/article_15014_4fa75dc06754787f27af2c7a4507c6e6.pdf