Derivative-free optimization for chemical product design

Sun, Yijia; Sahinidis, Nikolaos V.; Sundaram, Anantha; Cheon, Myun-Seok

doi:10.1016/j.coche.2019.11.006

Cited by 16 publications

(6 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Within these fixed procedures, different variants of local and global search methods are used, such as trust-region methods [68] and SNOBFIT [69]. The two subclasses of Derivative-Free Optimization (DFO) methods can be applied both directly without an underlying model, and indirectly in connection with a surrogate model that offers derivative approximations [70].…”

Section: Derivative-free Optimizationmentioning

confidence: 99%

See 1 more Smart Citation

Deep learning and knowledge-based methods for computer-aided molecular design—toward a unified approach: State-of-the-art and future directions

Alshehri

Gani²,

You

2020

Computers & Chemical Engineering

View full text Add to dashboard Cite

Section: Derivative-free Optimizationmentioning

confidence: 99%

“…However, it should be noted that the nature of DFO algorithms causes the quality of solutions to be highly dependent on the spaces and time limits, and hyperparameters that govern the search. Further, developed models tend to be only applicable to a specific class of molecular design problems [70], [80].…”

Section: Derivative-free Optimizationmentioning

confidence: 99%

Deep learning and knowledge-based methods for computer-aided molecular design—toward a unified approach: State-of-the-art and future directions

Alshehri

Gani²,

You

2020

Computers & Chemical Engineering

View full text Add to dashboard Cite

“…Recent reviews of DFO algorithms appear in Rios and Sahinidis (2013) 111 and Sun et al (2020). 112 In particular, the latter advocates the use of a portfolio of DFO algorithms, that is, a collection of DFO algorithms applied to the same problem, and demonstrates that this approach is more likely to produce good solutions in comparison to the application of a single algorithm. Especially appealing in this context is the ability of DFO algorithms to optimize in conjunction with collecting experimental measurements, thus potentially eliminating the need for predictive models.…”

Section: Materials Design For Organic Solar Cellsmentioning

confidence: 99%

Process Systems Engineering Perspective on the Design of Materials and Molecules

Adjiman

Sahinidis

Vlachos

et al. 2021

Ind. Eng. Chem. Res.

View full text Add to dashboard Cite

This paper aims to provide an overview of the current state of the art in molecule and material design and a perspective on open questions. First, we discuss the interplay between process design and the design of molecules and materials and then present the basic trade-offs and interdependencies that need to be considered in this integrated design problem. Second, we introduce methods and open questions in the area of processing materials, with special emphasis placed on heterogeneous catalysts due to their importance in the chemical industry. Finally, we discuss materials for photovoltaic cells as one example of a chemical product that is likely to help supply sustainable energy and address climate change concerns. What is presented here is based on presentations and discussions during the FIPSE-4 meeting in June 2018 (https://www.fi-in-pse.org/fipse-4-2018).

show abstract

“…However, we intend to check the performance of some deterministic algorithms on the same optimization problem, both to validate the results and to have a reference measure of the efficiency of the set of nature-inspired metaheuristic algorithms evaluated. According to Sun et al [47], when it is necessary to determine the optimum of a real-value function defined in n-dimensional space, but the derivatives are not available because they are computationally expensive or there is no explicit expression of the partial derivatives of the function, it is possible to resort to optimization methods called derivative-free or direct search methods. These methods use only function values and apply when no computer code can be produced for the derivative of the function.…”

Section: G Derivative-free Optimization Algorithmsmentioning

confidence: 99%

KPLS Optimization With Nature-Inspired Metaheuristic Algorithms

Mello-Román

Hernández

2020

IEEE Access

View full text Add to dashboard Cite

Kernel partial least squares regression (KPLS) is a technique used in several scientific areas because of its high predictive ability. This paper proposes a methodology to simultaneously estimate both the parameters of the kernel function and the number of components of the KPLS regression to maximize its predictive ability. A metaheuristic optimization problem was proposed taking the cumulative cross-validation coefficient as an objective function to be maximized. It was solved using nature-inspired metaheuristic algorithms: the genetic algorithm, particle swarm optimization, grey wolf optimization and the firefly algorithm. To validate the results and have a reference measure of the efficiency of the natureinspired metaheuristic algorithms, derivative-free optimization algorithms were also applied: Hooke-Jeeves and Nelder-Mead. The metaheuristic algorithms estimated optimal values of both of the kernel function parameters and the number of components in the KPLS regression. INDEX TERMS Partial least squares regression, kernel-based method, cross-validation method, and nature-inspired metaheuristic algorithms.

show abstract

Derivative-free optimization for chemical product design

Cited by 16 publications

References 54 publications

Deep learning and knowledge-based methods for computer-aided molecular design—toward a unified approach: State-of-the-art and future directions

Deep learning and knowledge-based methods for computer-aided molecular design—toward a unified approach: State-of-the-art and future directions

Process Systems Engineering Perspective on the Design of Materials and Molecules

KPLS Optimization With Nature-Inspired Metaheuristic Algorithms

Contact Info

Product

Resources

About