Sequential model based optimization of partially defined functions under unknown constraints

Abstract

This paper presents a sequential model based optimization framework for optimizing a black-box, multi-extremal and expensive objective function, which is also partially defined, that is it is undefined outside the feasible region. Furthermore, the constraints defining the feasible region within the search space are unknown. The approach proposed in this paper, namely SVM-CBO, is organized in two consecutive phases, the first uses a Support Vector Machine classifier to approximate the boundary of the unknown feasible region, the second uses Bayesian Optimization to find a globally optimal solution within the feasible region. In the first phase the next point to evaluate is chosen by dealing with the trade-off between improving the current estimate of the feasible region and discovering possible disconnected feasible sub-regions. In the second phase, the next point to evaluate is selected as the minimizer of the Lower Confidence Bound acquisition function but constrained to the current estimate of the feasible region. The main of the paper is a comparison with a Bayesian Optimization process which uses a fixed penalty value for infeasible function evaluations, under a limited budget (i.e., maximum number of function evaluations). Results are related to five 2D test functions from literature and 80 test functions, with increasing dimensionality and complexity, generated through the Emmental-type GKLS software. SVM-CBO proved to be significantly more effective as well as computationally efficient.

Lost in Optimization of Water Distribution Systems: Better Call Bayes

“…The hypervolume is an indicator of the solution quality that measures the objective space between a non-dominated set and a predefined reference vector. Figure 3 displays the mechanism of hypervolume improvement in the case of max 𝐹(𝑥) = (𝑓 (1) (𝑥), 𝑓 (2) (𝑥)). Figure 3a displays 4 points 𝑙 𝑖 with 𝑖 = 1, … ,4 non dominated in the objective space (the dominated space is in red), a new point 𝑥 1 dominates the previous ls, but 𝑙 1 bringing about an improvement of the dominate hypervolume (in blue, Figure 3b), x2 dominates 𝑙 1 ,𝑙 2 ,𝑙 3 and is not dominated by 𝑙 4 and brings about a further improvement (in green, Figure 3c).…”

“…A1 is mono-surrogate based on Chebyshev scalarization of the objectives of 𝑓 (1) and 𝑓 (2) . a GP of the aggregate function and the Expected Improvement (EI) as acquisition function.…”

SMGO-$Δ$: Balancing Caution and Reward in Global Optimization with Black-Box Constraints

“…Some works like [24,30] need feasible point/s in the initial data, but [14] addressed this issue by focusing on feasible region search when no feasible samples exist yet. [32] proposed a similar approach, devoting the initial iterations to learn the feasible regions, while [31] proposed alternating between exploitation (inside the feasible region) and exploration (discovering the constraint boundaries). Other works [33,34] leveraged in their sampling point selection the expected information gain w.r.t.…”

Optimization Method of Overall Parameters of Dual Pulse Missile Based on Surrogate Model

“…Bagheri [9] et al Proposed a new ego algorithm, which can well solve the constrained problems in the optimization problem and greatly reduce the calculation time cost. Antonio [10] et al Proposed an optimization framework based on sequence model (SVM-CBO) to optimize a black box, multi extremum and complex objective function. The multidisciplinary optimization methods mentioned above are difficult to balance their exploration and development ability, which is difficult to ensure the efficiency and global convergence of the optimization method.…”