Multi-Objective Bayesian Global Optimization using expected hypervolume improvement gradient

In multi-objective Bayesian optimization, an infill criterion is an important part, as it is the indicator to evaluate how much good a new set of solutions is, compared to a Pareto-front approximation set. This paper presents a deterministic algorithm for computing the Expected R2 Indicator for bi-objective problems and studies its use as an infill criterion in Bayesian Global Optimization. The R2-Indicator was introduced in 1998 by M. Hansen and A. Jaszkiewicz for performance assessment in multi-objective optimization and is more recently also used in indicator-based multi-criterion evolutionary algorithms (IBEAs). In Bayesian Global Optimization, we propose the Expected R2-indicator Improvement (ER2I) as an infill criterion. It is defined as the expected decrease of the R2 indicator by a point that is sampled from a predictive Gaussian distribution. The ER2I can also be used as a pre-selection criterion in surrogate-assisted IBEAs. It provides an alternative to the Expected Hypervolume-Indicator Improvement (EHVI) that requires a reference point, bounding the Pareto front from above. In contrast, the ER2I works with a utopian reference point that bounds the Pareto front from below. In addition, the ER2I supports preference modelling with utility functions and its computation time grows only linearly with the number of considered weight combinations. It is straightforward to approximate the ER2I by Monte Carlo Integration, but so far a deterministic algorithm to solve the non-linear integral remained unknown. We outline a deterministic algorithm for the computation of the bi-objective ER2I with Chebychev utility functions. Moreover, we study monotonicity properties of the ER2I w.r.t. parameters of the predictive distribution and numerical simulations demonstrate fast convergence to Pareto fronts of different shapes and the ability of the ER2I Bayesian optimization to fill gaps in the Pareto front approximation.

Multi-objective Bayesian optimization (MOBO) aims to optimize multiple competing objective functions in the expensive-to-evaluate scenario. The Expected Hypervolume Improvement (EHVI) is a commonly used acquisition function for MOBO and shows a good performance. However, the computation of EHVI becomes challenging as the number of objective functions grows. In this paper, we revisit the formulation of EHVI, as well as its multi-point counterpart qEHVI, and derive much simpler analytic expressions for them. The main contributions of this paper include: (1) first formulating EHVI as a particular hypervolume improvement, and thus immediately obtaining a formal proof of its NP-hardness, faster algorithms in both theory and practice, and more results on its derivatives; (2) first obtaining the analytic expressions of qEHVI for any q > 1 and m ≥ 2 where m is the number of objectives; and (3) demonstrating the advantages of our formulation over existing exact and approximation methods for computing EHVI and qEHVI through a large number of numerical experiments.

Surrogate-based method can dramatically reduce the number of expensive function evaluations in real-world multi- objective optimization problems (MOP). When the number of objectives is small, using surrogate models combined with expected hypervolume improvement (EHVI) infill sampling criterion (ISC) has been proved to be efficient to provide a set of solutions with good diversity and good proximity to the Pareto front (PF) in aerodynamic shape optimization. However, traditional hypervolume-based infilling strategies use only one kind of ISC to generate one or multiple sample points, the advantages of various kinds of ISC cannot be comprehensively utilized and the parallelization is not easy to implement. This paper proposes a combined multi-point infilling strategy based on Kriging models and develops an efficient global multi-objective constrained optimization method (EGMOCO) to solve multi- objective aerodynamic shape optimization with complex constraints. Multiple sample points are generated by using four ISC considering hypervolume at each iteration and then evaluated in parallel. Firstly, the performance of EGMOCO is compared with that of single criterion EHVI strategy on six benchmarks within the same computational budget to prove its effectiveness, and then EGMOCO is implemented in an aerodynamic shape optimization problem with complex constraints. The result shows that EGMOCO has good performance in balancing local exploitation and global exploration with faster convergence rate and high robustness, the whole PF can be fully explored in limited evaluations and the constraint handling is effective especially for real-world problems with complex and nonlinear constraints, the comprehensive aerodynamic performance of the airfoil is greatly improved. It can be confirmed that Kriging-based multi-objective optimization method combined with multi-point infilling strategy performs better than single infilling criterion EHVI, since different sample infilling criteria can complement with each other, both local exploitation and global exploration can be considered and well balanced.

Due to the adversarial limitations between the output solutions in the multi-objective optimization stage, the energy efficiency optimization of large pumping stations has the disadvantage of difficult coordination between power consumption and efficiency. Therefore, a multi-objective Bayesian hyperparameter optimization algorithm design for energy efficiency optimization of large pumping stations has been carried out. A multi-objective optimization model for energy efficiency of large pumping stations was constructed by combining the input power of the main motor, transmission energy loss, transformer loss, and energy consumption of auxiliary equipment, with efficiency as the constraint. We extract sequence features that affect pump water demand, and use feature labels as hyperparameters for the multi-objective optimization model of energy efficiency in large pumping stations. We transform the multi-objective optimization problem into a hyperparameter optimization problem, avoiding adversarial output solutions. We iterate the query points collected by EI, so the output model reaches a convergent optimal solution. The test results show that the designed algorithm can improve the pumping efficiency of the pump station while controlling the power consumption of the pump station, achieving collaborative optimization of efficiency.

We propose a surrogate-assisted reference vector adaptation (SRVA) method to\nsolve expensive multi- and many-objective optimization problems with various\nPareto front shapes. SRVA is coupled with a multi-objective Bayesian\noptimization (MBO) algorithm using reference vectors for scalarization of\nobjective functions. The Kriging surrogate models for MBO is used to estimate\nthe Pareto front shape and generate adaptive reference vectors uniformly\ndistributed on the estimated Pareto front. We combine SRVA with expected\nimprovement of penalty-based boundary intersection as an infill criterion for\nMBO. The proposed algorithm is compared with two other MBO algorithms by\napplying them to benchmark problems with various Pareto front shapes.\nExperimental results show that the proposed algorithm outperforms the other two\nin the problems whose objective functions are reasonably approximated by the\nKriging models. SRVA improves diversity of non-dominated solutions for these\nproblems with continuous, discontinuous, and degenerated Pareto fronts.\nBesides, the proposed algorithm obtains much better solutions from early stages\nof optimization especially in many-objective problems.\n

Bayesian optimization (BO) methods solve problems with several black-box objectives and constraints.Each black-box is expensive to evaluate and lacks a closed-form.They use a model of each black-box to guide the search for the problem's solution.Sometimes, however, the black-boxes may be evaluated at different fidelity levels.A lower fidelity is simply a cheap proxy of the corresponding black-box.Thus, lower fidelities that correlate with the actual black-box can be used to reduce the optimization cost.We propose Joint Entropy Search for Multi-Fidelity and Multi-objective Bayesian Optimization with Constraints (MF-JESMOC), a BO method for solving the aforementioned problems.It chooses the next point and fidelity level at which to evaluate the black-boxes as the one that is expected to reduce the most the joint entropy of the Pareto set and the Pareto front, normalized by the fidelity's cost.Deep Gaussian processes are used to model each black-box and dependencies between fidelities.In our experiments, MF-JESMOC outperforms other state-of-the-art methods for multi-objective BO with constraints and different fidelity levels.

Optimizing multiple conflicting black-box objectives simultaneously is a prevalent occurrence in many real-world applications, such as neural architecture search, and machine learning. These problems are known as expensive multi-objective optimization problems (EMOPs) when the function evaluations are computationally or financially costly. Multi-objective Bayesian optimization (MOBO) offers an efficient approach to discovering a set of Pareto optimal solutions. However, the data deficiency issue caused by limited function evaluations has posed a great challenge to current optimization methods. Moreover, most current methods tend to prioritize the quality of candidate solutions, while ignoring the quantity of promising samples. In order to tackle these issues, our paper proposes a novel multi-objective Bayesian optimization algorithm with a data augmentation strategy that provides ample high-quality samples for Pareto set learning (PSL). Specifically, it utilizes Generative Adversarial Networks (GANs) to enrich data and a dominance prediction model to screen out high-quality samples, mitigating the predicament of limited function evaluations in EMOPs. Additionally, we adopt the regularity model to expensive multi-objective Bayesian optimization for PSL. Experimental results on both synthetic and real-world problems demonstrate that our algorithm outperforms several state-of-the-art and classical algorithms.

Rather than optimizing expensive objective functions such as complex engineering simulations directly, Bayesian optimization methodologies fit a surrogate model (typically Kriging or a Gaussian Process) on evaluations of the objective function(s). To determine the next evaluation, an acquisition function is optimized (also referred to as infill criterion or sampling policy) which incorporates the model prediction and uncertainty and balances exploration and exploitation. Therefore, Bayesian optimization methodologies replace a single optimization of the objective function by a sequence of optimization problems: this makes sense as the acquisition function is cheap-to-evaluate whereas the objective is not. Depending on the goal different acquisition functions are available: multi-objective acquisition functions are relatively new and this chapter gives a state-of-the-art overview and illustrates some approaches based on hypervolume improvement. It is shown that the quality of the model is crucial for the performance of Bayesian optimization and illustrate this by using the more flexible Student-t processes as surrogate models.

The traditional multi-objective efficient global optimization (EGO) algorithms have been hybridized and adapted to solving the expensive aerodynamic shape optimization problems based on high-fidelity numerical simulations. Although the traditional EGO algorithms are highly efficient in solving some of the optimization problems with very complex landscape, it is not preferred to solve most of the aerodynamic shape optimization problems with relatively low-degree multi-modal design spaces. A new infill criterion encouraging more local exploitation has been proposed by hybridizing two traditional multi-objective expected improvements (EIs), namely, statistical multi-objective EI and expected hypervolume improvement, in order to improve their robustness and efficiency in aerodynamic shape optimization. Different analytical test problems and aerodynamic shape optimization problems have been investigated. In comparison with traditional multi-objective EI algorithms and a standard evolutionary multi-objective optimization algorithm, the proposed method is shown to be more robust and efficient in the tests due to its hybrid characteristics, easier handling of sub-optimization problems, and enhanced exploitation capability.

In the aircraft design methodology, there are many methods to measuring the UAVs performance, such as, the minimize drag, the lift required or the take-off/landing performance etc. The multi-objective optimization is one of the popular method for UAVs design problem. In addition, in the UAVs design problem is required for high computational cost such as computation of fluid dynamics. The solution to the said problem could be reached through the method of the Efficient Global Optimization algorithm (or abbreviated, the EGO). However, the EGO was, in the first instant, intended for very limited use; namely, it was utilized as a solution for a single-objective optimization problem with just one additional sampling. Then, the EGO method must be required for long computational time for single-additional sampling procedure. The objective of this particular research was to study the EGO with multi-objective multi-additional sampling as a solution to the UAVs airfoil design problem. The Expected Hypervolume Improvement (EHVI) is applied with the EGO process with an intention to find a solution to the multi-objective optimization problem. Furthermore, there was a proposal to use multiple additional sampling methods in the efficiency improvement of the additional sampling process in EGO, and at the same time, keeping the performance of exploration based on EHVI maximization at the same constant There are two main goals in the application of this algorithm to UAVs airfoil design optimization, which include minimizing aerodynamic drag and maximizing UAVs airfoil thickness at the trailing edge. The Reynolds-averaged Navier-Stokes simulation is applied for aerodynamic evaluation. By adopting the airfoil design, the results were the reduction in the aerodynamic drag, as well as 5% improvement of the thickness of the airfoil at the trailing edge when compared with the airfoil initial design.

We report a human-in-the-loop implementation of the multi-objective experimental design via a Bayesian optimization platform (EDBO+) towards the optimization of butylpyridinium bromide synthesis under continuous flow conditions. The algorithm simultaneously optimized reaction yield and production rate (or space-time yield) and generated a well defined Pareto front. The versatility of EDBO+ was demonstrated by expanding the reaction space mid-campaign by increasing the upper temperature limit. Incorporation of continuous flow techniques enabled improved control over reaction parameters compared to common batch chemistry processes, while providing a route towards future automated syntheses and improved scalability. To that end, we applied the open-source Python module, nmrglue, for semi-automated nuclear magnetic resonance (NMR) spectroscopy analysis, and compared the acquired outputs against those obtained through manual processing methods from spectra collected on both low-field (60 MHz) and high-field (400 MHz) NMR spectrometers. The EDBO+ based model was retrained with these four different datasets and the resulting Pareto front predictions provided insight into the effect of data analysis on model predictions. Finally, quaternization of poly(4-vinylpyridine) with bromobutane illustrated the extension of continuous flow chemistry to synthesize functional materials.

Global Sensitivity Analysis (GSA) is the study of the influence of any given input on the outputs of a model. In the context of engineering design, GSA has been widely used to understand both individual and collective contributions of design variables on the design objectives. So far, global sensitivity studies have often been limited to design spaces with only quantitative (numerical) design variables. However, many engineering systems also contain, if not only, qualitative (categorical) design variables in addition to quantitative design variables. In this paper, we integrate Latent Variable Gaussian Process (LVGP) with Sobol’ analysis to develop the first metamodel-based mixed-variable GSA method. Through numerical case studies, we validate and demonstrate the effectiveness of our proposed method for mixed-variable problems. Furthermore, while the proposed GSA method is general enough to benefit various engineering design applications, we integrate it with multi-objective Bayesian optimization (BO) to create a sensitivity-aware design framework in accelerating the Pareto front design exploration for metal-organic framework (MOF) materials with many-level combinatorial design spaces. Although MOFs are constructed only from qualitative variables that are notoriously difficult to design, our method can utilize sensitivity analysis to navigate the optimization in the many-level large combinatorial design space, greatly expediting the exploration of novel MOF candidates.

Wireless cellular networks have many parameters that are normally tuned upon deployment and re-tuned as the network changes. Many operational parameters affect reference signal received power (RSRP), reference signal received quality (RSRQ), signal-to-interference-plus-noise-ratio (SINR), and, ultimately, throughput. In this paper, we develop and compare two approaches for maximizing coverage and minimizing interference by jointly optimizing the transmit power and downtilt (elevation tilt) settings across sectors. To evaluate different parameter configurations offline, we construct a realistic simulation model that captures geographic correlations. Using this model, we evaluate two optimization methods: deep deterministic policy gradient (DDPG), a reinforcement learning (RL) algorithm, and multi-objective Bayesian optimization (BO). Our simulations show that both approaches significantly outperform random search and converge to comparable Pareto frontiers, but that BO converges with two orders of magnitude fewer evaluations than DDPG. Our results suggest that data-driven techniques can effectively self-optimize coverage and capacity in cellular networks.

Global Sensitivity Analysis (GSA) is the study of the influence of any given inputs on the outputs of a model. In the context of engineering design, GSA has been widely used to understand both individual and collective contributions of design variables on the design objectives. So far, global sensitivity studies have often been limited to design spaces with only quantitative (numerical) design variables. However, many engineering systems also contain, if not only, qualitative (categorical) design variables in addition to quantitative design variables. In this paper, we integrate the novel Latent Variable Gaussian Process (LVGP) with Sobol’ analysis to develop the first metamodel-based mixed-variable GSA method. Through two analytical case studies, we first validate and demonstrate the effectiveness of our proposed method for mixed-variable problems. Furthermore, while the new metamodel-based mixed-variable GSA method can benefit various engineering design applications, we employ our method with multi-objective Bayesian optimization (BO) to accelerate the Pareto front design exploration in many-level combinatorial design spaces. Specifically, we implement a sensitivity-aware design framework for metal-organic framework (MOF) materials that are constructed only from qualitative design variables and show the benefits of our method for expediting the exploration of novel MOF candidates from a many-level large combinatorial design space.

Start-to-end modeling and transmission efficiency optimization for a cyclotron-based proton therapy beamline