scholarly journals Extending Expected Improvement for High-Dimensional Stochastic Optimization of Expensive Black-Box Functions

2016 ◽  
Author(s):  
Piyush Pandita ◽  
Ilias Bilionis ◽  
Jitesh Panchal
Keyword(s):  
Global Optimization ◽  
Stochastic Optimization ◽  
Objective Function ◽  
Design Optimization ◽  
Information Acquisition ◽  
Optimization Problems ◽  
Epistemic Uncertainty ◽  
Optimization Under Uncertainty ◽  
Expected Improvement ◽  
Convergence Criterion

Design optimization under uncertainty is notoriously difficult when the objective function is expensive to evaluate. State-of-the-art techniques, e.g., stochastic optimization or sampling average approximation, fail to learn exploitable patterns from collected data and, as a result, they tend to require an excessive number of objective function evaluations. There is a need for techniques that alleviate the high cost of information acquisition and select sequential simulations in an optimal way. In the field of deterministic single-objective unconstrained global optimization, the Bayesian global optimization (BGO) approach has been relatively successful in addressing the information acquisition problem. BGO builds a probabilistic surrogate of the expensive objective function and uses it to define an information acquisition function (IAF) whose role is to quantify the merit of making new objective evaluations. Specifically, BGO iterates between making the observations with the largest expected IAF and rebuilding the probabilistic surrogate, until a convergence criterion is met. In this work, we extend the expected improvement (EI) IAF to the case of design optimization under uncertainty. This involves a reformulation of the EI policy that is able to filter out parametric and measurement uncertainties. We by-pass the curse of dimensionality, since the method does not require learning the response surface as a function of the stochastic parameters. To increase the robustness of our approach in the low sample regime, we employ a fully Bayesian interpretation of Gaussian processes by constructing a particle approximation of the posterior of its hyperparameters using adaptive Markov chain Monte Carlo. An addendum of our approach is that it can quantify the epistemic uncertainty on the location of the optimum and the optimal value as induced by the limited number of objective evaluations used in obtaining it. We verify and validate our approach by solving two synthetic optimization problems under uncertainty. We demonstrate our approach by solving a challenging engineering problem: the oil-well-placement problem with uncertainties in the permeability field and the oil price time series.

scholarly journals Extending Expected Improvement for High-Dimensional Stochastic Optimization of Expensive Black-Box Functions

2016 ◽  
Vol 138 (11) ◽  
Author(s):  
Piyush Pandita ◽  
Ilias Bilionis ◽  
Jitesh Panchal
Keyword(s):  
Global Optimization ◽  
Stochastic Optimization ◽  
Objective Function ◽  
Information Acquisition ◽  
Optimization Problems ◽  
Epistemic Uncertainty ◽  
Optimization Under Uncertainty ◽  
Expected Improvement ◽  
Oil Well ◽  
Placement Problem

Design optimization under uncertainty is notoriously difficult when the objective function is expensive to evaluate. State-of-the-art techniques, e.g., stochastic optimization or sampling average approximation, fail to learn exploitable patterns from collected data and require a lot of objective function evaluations. There is a need for techniques that alleviate the high cost of information acquisition and select sequential simulations optimally. In the field of deterministic single-objective unconstrained global optimization, the Bayesian global optimization (BGO) approach has been relatively successful in addressing the information acquisition problem. BGO builds a probabilistic surrogate of the expensive objective function and uses it to define an information acquisition function (IAF) that quantifies the merit of making new objective evaluations. In this work, we reformulate the expected improvement (EI) IAF to filter out parametric and measurement uncertainties. We bypass the curse of dimensionality, since the method does not require learning the response surface as a function of the stochastic parameters, and we employ a fully Bayesian interpretation of Gaussian processes (GPs) by constructing a particle approximation of the posterior of its hyperparameters using adaptive Markov chain Monte Carlo (MCMC) to increase the methods robustness. Also, our approach quantifies the epistemic uncertainty on the location of the optimum and the optimal value as induced by the limited number of objective evaluations used in obtaining it. We verify and validate our approach by solving two synthetic optimization problems under uncertainty and demonstrate it by solving the oil-well placement problem (OWPP) with uncertainties in the permeability field and the oil price time series.

A hybrid global optimization method based on multiple metamodels

2018 ◽  
Vol 35 (1) ◽  
pp. 71-90 ◽  
Author(s):  
Xiwen Cai ◽  
Haobo Qiu ◽  
Liang Gao ◽  
Xiaoke Li ◽  
Xinyu Shao
Keyword(s):  
Global Optimization ◽  
Design Optimization ◽  
Engineering Design ◽  
Optimization Problems ◽  
Optimization Methods ◽  
Optimization Method ◽  
Expected Improvement ◽  
Benchmark Problems ◽  
Content Type ◽  
Engineering Design Optimization

Purpose This paper aims to propose hybrid global optimization based on multiple metamodels for improving the efficiency of global optimization. Design/methodology/approach The method has fully utilized the information provided by different metamodels in the optimization process. It not only imparts the expected improvement criterion of kriging into other metamodels but also intelligently selects appropriate metamodeling techniques to guide the search direction, thus making the search process very efficient. Besides, the corresponding local search strategies are also put forward to further improve the optimizing efficiency. Findings To validate the method, it is tested by several numerical benchmark problems and applied in two engineering design optimization problems. Moreover, an overall comparison between the proposed method and several other typical global optimization methods has been made. Results show that the global optimization efficiency of the proposed method is higher than that of the other methods for most situations. Originality/value The proposed method sufficiently utilizes multiple metamodels in the optimizing process. Thus, good optimizing results are obtained, showing great applicability in engineering design optimization problems which involve costly simulations.

scholarly journals Global optimization based on active preference learning with radial basis functions

2020 ◽  
Author(s):  
Alberto Bemporad ◽  
Dario Piga
Keyword(s):  
Global Optimization ◽  
Objective Function ◽  
Decision Maker ◽  
Optimization Problems ◽  
Global Optimum ◽  
Bayesian Optimization ◽  
Preference Learning ◽  
Global Optimizer ◽  
Radial Basis ◽  
Decision Vector

AbstractThis paper proposes a method for solving optimization problems in which the decision-maker cannot evaluate the objective function, but rather can only express a preference such as “this is better than that” between two candidate decision vectors. The algorithm described in this paper aims at reaching the global optimizer by iteratively proposing the decision maker a new comparison to make, based on actively learning a surrogate of the latent (unknown and perhaps unquantifiable) objective function from past sampled decision vectors and pairwise preferences. A radial-basis function surrogate is fit via linear or quadratic programming, satisfying if possible the preferences expressed by the decision maker on existing samples. The surrogate is used to propose a new sample of the decision vector for comparison with the current best candidate based on two possible criteria: minimize a combination of the surrogate and an inverse weighting distance function to balance between exploitation of the surrogate and exploration of the decision space, or maximize a function related to the probability that the new candidate will be preferred. Compared to active preference learning based on Bayesian optimization, we show that our approach is competitive in that, within the same number of comparisons, it usually approaches the global optimum more closely and is computationally lighter. Applications of the proposed algorithm to solve a set of benchmark global optimization problems, for multi-objective optimization, and for optimal tuning of a cost-sensitive neural network classifier for object recognition from images are described in the paper. MATLAB and a Python implementations of the algorithms described in the paper are available at http://cse.lab.imtlucca.it/~bemporad/glis.

scholarly journals On quantitative stability in infinite-dimensional optimization under uncertainty

2021 ◽  
Author(s):  
M. Hoffhues ◽  
W. Römisch ◽  
T. M. Surowiec
Keyword(s):  
Stochastic Optimization ◽  
Constrained Optimization ◽  
Optimization Problems ◽  
Optimization Under Uncertainty ◽  
Optimal Solutions ◽  
Probability Metrics ◽  
Pde Constrained Optimization ◽  
Quantitative Stability ◽  
Infinite Dimensional ◽  
Optimal Value

AbstractThe vast majority of stochastic optimization problems require the approximation of the underlying probability measure, e.g., by sampling or using observations. It is therefore crucial to understand the dependence of the optimal value and optimal solutions on these approximations as the sample size increases or more data becomes available. Due to the weak convergence properties of sequences of probability measures, there is no guarantee that these quantities will exhibit favorable asymptotic properties. We consider a class of infinite-dimensional stochastic optimization problems inspired by recent work on PDE-constrained optimization as well as functional data analysis. For this class of problems, we provide both qualitative and quantitative stability results on the optimal value and optimal solutions. In both cases, we make use of the method of probability metrics. The optimal values are shown to be Lipschitz continuous with respect to a minimal information metric and consequently, under further regularity assumptions, with respect to certain Fortet-Mourier and Wasserstein metrics. We prove that even in the most favorable setting, the solutions are at best Hölder continuous with respect to changes in the underlying measure. The theoretical results are tested in the context of Monte Carlo approximation for a numerical example involving PDE-constrained optimization under uncertainty.

scholarly journals A stability result for linear Markovian stochastic optimization problems

2020 ◽  
Author(s):  
Adriana Kiszka ◽  
David Wozabal
Keyword(s):  
Stochastic Optimization ◽  
Stochastic Processes ◽  
Objective Function ◽  
Markov Processes ◽  
Optimization Problems ◽  
Stability Result ◽  
Extant Literature ◽  
Numerical Example ◽  
Optimal Values ◽  
Objective Value

Abstract In this paper, we propose a semi-metric for Markov processes that allows to bound optimal values of linear Markovian stochastic optimization problems. Similar to existing notions of distance for general stochastic processes, our distance is based on transportation metrics. As opposed to the extant literature, the proposed distance is problem specific, i.e., dependent on the data of the problem whose objective value we want to bound. As a result, we are able to consider problems with randomness in the constraints as well as in the objective function and therefore relax an assumption in the extant literature. We derive several properties of the proposed semi-metric and demonstrate its use in a stylized numerical example.

Expected-Improvement-Based Methods for Adaptive Sampling in Multi-Objective Optimization Problems

2016 ◽  
Author(s):  
Jesper Kristensen ◽  
You Ling ◽  
Isaac Asher ◽  
Liping Wang
Keyword(s):  
Convex Hull ◽  
Design Optimization ◽  
Adaptive Sampling ◽  
Pareto Front ◽  
Performance Metrics ◽  
Sampling Methods ◽  
Optimization Problems ◽  
Expected Improvement ◽  
Multi Objective Optimization ◽  
Multi Objective

Adaptive sampling methods have been used to build accurate meta-models across large design spaces from which engineers can explore data trends, investigate optimal designs, study the sensitivity of objectives on the modeling design features, etc. For global design optimization applications, adaptive sampling methods need to be extended to sample more efficiently near the optimal domains of the design space (i.e., the Pareto front/frontier in multi-objective optimization). Expected Improvement (EI) methods have been shown to be efficient to solve design optimization problems using meta-models by incorporating prediction uncertainty. In this paper, a set of state-of-the-art methods (hypervolume EI method and centroid EI method) are presented and implemented for selecting sampling points for multi-objective optimizations. The classical hypervolume EI method uses hyperrectangles to represent the Pareto front, which shows undesirable behavior at the tails of the Pareto front. This issue is addressed utilizing the concepts from physical programming to shape the Pareto front. The modified hypervolume EI method can be extended to increase local Pareto front accuracy in any area identified by an engineer, and this method can be applied to Pareto frontiers of any shape. A novel hypervolume EI method is also developed that does not rely on the assumption of hyperrectangles, but instead assumes the Pareto frontier can be represented by a convex hull. The method exploits fast methods for convex hull construction and numerical integration, and results in a Pareto front shape that is desired in many practical applications. Various performance metrics are defined in order to quantitatively compare and discuss all methods applied to a particular 2D optimization problem from the literature. The modified hypervolume EI methods lead to dramatic resource savings while improving the predictive capabilities near the optimal objective values.

Hybrid and Adaptive Metamodel Based Global Optimization

2009 ◽  
Author(s):  
J. Gu ◽  
G. Y. Li ◽  
Z. Dong
Keyword(s):  
Global Optimization ◽  
Design Optimization ◽  
Optimization Problems ◽  
Optimization Method ◽  
Global Optimum ◽  
Superior Performance ◽  
Automatic Identification ◽  
Design Problems ◽  
Sample Data ◽  
Data Points

Metamodeling techniques are increasingly used in solving computation intensive design optimization problems today. In this work, the issue of automatic identification of appropriate metamodeling techniques in global optimization is addressed. A generic, new hybrid metamodel based global optimization method, particularly suitable for design problems involving computation intensive, black-box analyses and simulations, is introduced. The method employs three representative metamodels concurrently in the search process and selects sample data points adaptively according to the values calculated using the three metamodels to improve the accuracy of modeling. The global optimum is identified when the metamodels become reasonably accurate. The new method is tested using various benchmark global optimization problems and applied to a real industrial design optimization problem involving vehicle crash simulation, to demonstrate the superior performance of the new algorithm over existing search methods. Present limitations of the proposed method are also discussed.

scholarly journals Global optimization of laminated composite beams using an improved differential evolution algorithm

2020 ◽  
Vol 14 (1) ◽  
pp. 54-64
Author(s):  
Lam Thuan Phat ◽  
Nguyen Nhat Phi Long ◽  
Nguyen Hoai Son ◽  
Ho Huu Vinh ◽  
Le Anh Thang
Keyword(s):  
Global Optimization ◽  
Differential Evolution ◽  
Design Optimization ◽  
Optimization Problems ◽  
Differential Evolution Algorithm ◽  
Laminated Composite ◽  
Roulette Wheel Selection ◽  
Roulette Wheel ◽  
Evolution Algorithm ◽  
Improved Differential Evolution Algorithm

Differential Evolution (DE) is an efficient and effective algorithm recently proposed for solving optimization problems. In this paper, an improved version of Differential Evolution algorithm, called iDE, is introduced to solve design optimization problems of composite laminated beams. The beams used in this research are Timoshenko beam models computed based on analytical formula. The iDE is formed by modifying the mutation and the selection step of the original algorithm. Particularly, individuals involved in mutation were chosen by Roulette wheel selection via acceptant stochastic instead of the random selection. Meanwhile, in selection phase, the elitist operator is used for the selection progress instead of basic selection in the optimization process of the original DE algorithm. The proposed method is then applied to solve two problems of lightweight design optimization of the Timoshenko laminated composite beam with discrete variables. Numerical results obtained have been compared with those of the references and proved the effectiveness and efficiency of the proposed method. Keywords: improved Differential Evolution algorithm; Timoshenko composite laminated beam; elitist operator; Roulette wheel selection; deterministic global optimization.

scholarly journals Aggregation–Decomposition-Based Multi-Agent Reinforcement Learning for Multi-Reservoir Operations Optimization

Water ◽  
2020 ◽  
Vol 12 (10) ◽  
pp. 2688 ◽  
Author(s):  
Milad Hooshyar ◽  
S. Jamshid Mousavi ◽  
Masoud Mahootchi ◽  
Kumaraswamy Ponnambalam
Keyword(s):  
Reinforcement Learning ◽  
Stochastic Optimization ◽  
Optimization Problems ◽  
Transition Probability ◽  
Optimization Under Uncertainty ◽  
Stochastic Dynamic ◽  
Optimization Approach ◽  
Reservoir Operations ◽  
Multi Agent ◽  
Operations Optimization

Stochastic dynamic programming (SDP) is a widely-used method for reservoir operations optimization under uncertainty but suffers from the dual curses of dimensionality and modeling. Reinforcement learning (RL), a simulation-based stochastic optimization approach, can nullify the curse of modeling that arises from the need for calculating a very large transition probability matrix. RL mitigates the curse of the dimensionality problem, but cannot solve it completely as it remains computationally intensive in complex multi-reservoir systems. This paper presents a multi-agent RL approach combined with an aggregation/decomposition (AD-RL) method for reducing the curse of dimensionality in multi-reservoir operation optimization problems. In this model, each reservoir is individually managed by a specific operator (agent) while co-operating with other agents systematically on finding a near-optimal operating policy for the whole system. Each agent makes a decision (release) based on its current state and the feedback it receives from the states of all upstream and downstream reservoirs. The method, along with an efficient artificial neural network-based robust procedure for the task of tuning Q-learning parameters, has been applied to a real-world five-reservoir problem, i.e., the Parambikulam–Aliyar Project (PAP) in India. We demonstrate that the proposed AD-RL approach helps to derive operating policies that are better than or comparable with the policies obtained by other stochastic optimization methods with less computational burden.

Sign in / Sign up

Export Citation Format

Share Document