0000000000040339
AUTHOR
Atanu Mazumdar
Treed Gaussian Process Regression for Solving Offline Data-Driven Continuous Multiobjective Optimization Problems
Abstract For offline data-driven multiobjective optimization problems (MOPs), no new data is available during the optimization process. Approximation models (or surrogates) are first built using the provided offline data and an optimizer, e.g. a multiobjective evolutionary algorithm, can then be utilized to find Pareto optimal solutions to the problem with surrogates as objective functions. In contrast to online data-driven MOPs, these surrogates cannot be updated with new data and, hence, the approximation accuracy cannot be improved by considering new data during the optimization process. Gaussian process regression (GPR) models are widely used as surrogates because of their ability to pr…
An Interactive Framework for Offline Data-Driven Multiobjective Optimization
We propose a framework for solving offline data-driven multiobjective optimization problems in an interactive manner. No new data becomes available when solving offline problems. We fit surrogate models to the data to enable optimization, which introduces uncertainty. The framework incorporates preference information from a decision maker in two aspects to direct the solution process. Firstly, the decision maker can guide the optimization by providing preferences for objectives. Secondly, the framework features a novel technique for the decision maker to also express preferences related to maximum acceptable uncertainty in the solutions as preferred ranges of uncertainty. In this way, the d…
Probabilistic Selection Approaches in Decomposition-based Evolutionary Algorithms for Offline Data-Driven Multiobjective Optimization
In offline data-driven multiobjective optimization, no new data is available during the optimization process. Approximation models, also known as surrogates, are built using the provided offline data. A multiobjective evolutionary algorithm can be utilized to find solutions by using these surrogates. The accuracy of the approximated solutions depends on the surrogates and approximations typically involve uncertainties. In this paper, we propose probabilistic selection approaches that utilize the uncertainty information of the Kriging models (as surrogates) to improve the solution process in offline data-driven multiobjective optimization. These approaches are designed for decomposition-base…
Optimistic NAUTILUS navigator for multiobjective optimization with costly function evaluations
AbstractWe introduce novel concepts to solve multiobjective optimization problems involving (computationally) expensive function evaluations and propose a new interactive method called O-NAUTILUS. It combines ideas of trade-off free search and navigation (where a decision maker sees changes in objective function values in real time) and extends the NAUTILUS Navigator method to surrogate-assisted optimization. Importantly, it utilizes uncertainty quantification from surrogate models like Kriging or properties like Lipschitz continuity to approximate a so-called optimistic Pareto optimal set. This enables the decision maker to search in unexplored parts of the Pareto optimal set and requires …
Interactive evolutionary multiobjective optimization with modular physical user interface
© 2022 Copyright held by the owner/author(s). Incorporating the preferences of a domain expert, a decision-maker (DM), in solving multiobjective optimization problems increased in popularity in recent years. The DM can choose to use different types of preferences depending on his/her comfort, requirements, or the problem being solved. Most papers, where preference-based and interactive algorithms have been proposed, do not pay attention to the user interfaces and input devices. If they do, they use character or graphics-based preference input methods. We propose the option of using a physical or tactile input device that gives the DM a better sense of control over providing his/her preferen…
On Dealing with Uncertainties from Kriging Models in Offline Data-Driven Evolutionary Multiobjective Optimization
Many works on surrogate-assisted evolutionary multiobjective optimization have been devoted to problems where function evaluations are time-consuming (e.g., based on simulations). In many real-life optimization problems, mathematical or simulation models are not always available and, instead, we only have data from experiments, measurements or sensors. In such cases, optimization is to be performed on surrogate models built on the data available. The main challenge there is to fit an accurate surrogate model and to obtain meaningful solutions. We apply Kriging as a surrogate model and utilize corresponding uncertainty information in different ways during the optimization process. We discuss…