Grant: Discovery Grants Program - Individual
École Polytechnique de Montréal
$50,000.00
Blackbox optimization: Algorithms, software, and applications
FY 2024-25
Summary
The focus of my research is on blackbox optimization, where problems are typically provided through computer experiments. Such problems are diverse and prevalent across all areas of Engineering and beyond. These real-world applications often come from industry or applied research laboratories, as the blackbox approach is often the only practical way to accurately model these scenarios. Traditional methods being not applicable due to the absence of derivative information, I consider derivative-free optimization methods such as the Mesh Adaptive Direct Search algorithm (MADS). Over the last 12 years covered by my two previous Discovery Grants, I have contributed to MADS and its open-source software, NOMAD. This proposal represents the direct continuation of this successful research. It outlines a five-year research program centered on extensions of MADS, with a focus on algorithm design and analysis, software development, and applications in diverse fields, with an emphasis on machine learning and benchmarking. The program will directly impact the optimization community and benefit Canada as it keeps a cutting-edge expertise in Montreal and contributes to the training of 12 students with advanced skills in optimization. It is divided into 10 projects that will touch upon the following categories:
Algorithms: MADS still has certain limitations, notably in addressing large-scale problems. Two of my projects aim to overcome this limitation. The Proposal also includes new variations of MADS, further handling of constraints, and new strategies to handle categorical variables. Another project involves surrogates with controllable multi-fidelity. Finally, I will consider greyboxes, i.e. blackboxes for which partial structure is known. All the new features will be mathematically analyzed to ensure the availability of convergence results.
Software: NOMAD is continuously evolving to integrate the latest algorithmic developments. Recently, NOMAD was entirely recoded to be more flexible and adaptable and is now natively parallel. With its solid user base, many features, robustness, and efficiency, NOMAD has established itself as a baseline in blackbox optimization. While successful algorithmic improvements made possible by this research will be integrated into NOMAD, there will also be other significant contributions in terms of software since several blackbox applications will be made available to the community.
Applications: They drive this research, serving as both motivation and validation for the developed methods. My most recent problems encompass a chemical process simulation, a fully configurable solar power plant, and a collection of multidisciplinary optimization problems. Thanks to industrial and academic partners, I have access to a wide range of metaheuristics and neural networks hyperparameters tuning problems, and applications in aeronautics. Whenever possible, my objective is to distribute these applications for the optimization community.
Program
Discovery Grants Program - Individual
Recipient
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2024-05086
Name nom
Le Digabel, Sébastien
Department département
MAGI
Organization id
64
Institution établissement
École Polytechnique de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
50000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1509
Committee name en
Civil, Industrial and Systems Engineering
Committee name fr
Génie civil, industriel et des systèmes
Area of application code
1200
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Advancement of knowledge
Area of application fr
Avancement des connaissances
Research subject code
2956
Research subject group en
Applied mathematics
Research subject group fr
Mathématiques appliquées
Research subject en
Optimisation and optimal control theory
Research subject fr
Optimisation et théories de commande optimale
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Blackbox optimization: Algorithms, software, and applications
Keywords
mathematical optimization; derivative-free optimization; blackbox optimization; optimization algorithms; algorithms benchmarking; realistic applications
Application summary
The focus of my research is on blackbox optimization, where problems are typically provided through computer experiments. Such problems are diverse and prevalent across all areas of Engineering and beyond. These real-world applications often come from industry or applied research laboratories, as the blackbox approach is often the only practical way to accurately model these scenarios. Traditional methods being not applicable due to the absence of derivative information, I consider derivative-free optimization methods such as the Mesh Adaptive Direct Search algorithm (MADS). Over the last 12 years covered by my two previous Discovery Grants, I have contributed to MADS and its open-source software, NOMAD. This proposal represents the direct continuation of this successful research. It outlines a five-year research program centered on extensions of MADS, with a focus on algorithm design and analysis, software development, and applications in diverse fields, with an emphasis on machine learning and benchmarking. The program will directly impact the optimization community and benefit Canada as it keeps a cutting-edge expertise in Montreal and contributes to the training of 12 students with advanced skills in optimization. It is divided into 10 projects that will touch upon the following categories:
Algorithms: MADS still has certain limitations, notably in addressing large-scale problems. Two of my projects aim to overcome this limitation. The Proposal also includes new variations of MADS, further handling of constraints, and new strategies to handle categorical variables. Another project involves surrogates with controllable multi-fidelity. Finally, I will consider greyboxes, i.e. blackboxes for which partial structure is known. All the new features will be mathematically analyzed to ensure the availability of convergence results.
Software: NOMAD is continuously evolving to integrate the latest algorithmic developments. Recently, NOMAD was entirely recoded to be more flexible and adaptable and is now natively parallel. With its solid user base, many features, robustness, and efficiency, NOMAD has established itself as a baseline in blackbox optimization. While successful algorithmic improvements made possible by this research will be integrated into NOMAD, there will also be other significant contributions in terms of software since several blackbox applications will be made available to the community.
Applications: They drive this research, serving as both motivation and validation for the developed methods. My most recent problems encompass a chemical process simulation, a fully configurable solar power plant, and a collection of multidisciplinary optimization problems. Thanks to industrial and academic partners, I have access to a wide range of metaheuristics and neural networks hyperparameters tuning problems, and applications in aeronautics. Whenever possible, my objective is to distribute these applications for the optimization community.