Grant: Discovery Grants Program - Individual
École Polytechnique de Montréal
$52,000.00
Interpretable Decision Processes in Data-Driven Supply Chains
EF 2024-25
Résumé
Supply chains are under constant transformation as a consequence of disruptive events (e.g., COVID-19), shifts in consumer behavior and new transportation paradigms (e.g., crowdsourced deliveries, autonomous vehicles). The greater availability of data means that more sophisticated decision and control processes can be deployed in practice. However, as these algorithms become more widespread and complex, new concerns arise concerning decision transparency and reliability. As of today, the successful application of machine learning and optimization algorithms rests on a delicate equilibrium: while the methods should be as sophisticated as required, they should also remain as simple and interpretable as possible to achieve fairness and accountability.
To address these important challenges, this research program focuses on the aspects of algorithmic simplicity, explainability, and robustness in data-driven supply chains and production systems. Our goal is to progress towards decision-support algorithms that are competitive in terms of economic performance, but that also possess the qualities mentioned above. To achieve these goals, we pursue cross-disciplinary methodological studies building upon machine learning techniques (e.g., decision tree-based predictors) and combinatorial optimization algorithms (e.g., metaheuristics and mathematical programming algorithms).
� We start our research by integrating explainable machine learning algorithms in transportation and production optimization. We use the capabilities of machine learning algorithms in contexts where it is difficult to estimate the impact of the decisions efficiently, either due to sources of uncertainty (stochastic and robust optimization) or long planning horizons (strategic optimization). We aim to learn decision rules to 1) prevent and react to disruptions and 2) to learn the economic impact of strategic decisions, and embed these components in state-of-the-art optimization algorithms.
� In the second part of this program, we refine available machine learning models and algorithms to adequate them to our applications and improve their performance, using combinatorial optimization techniques. We will pursue our research on model compression (i.e., simplification of a learned machine learning model) for decision-tree and neural-network classifiers. Moreover, we use combinatorial optimization algorithms to produce trustworthy "counterfactual explanations," i.e., smallest perturbations of an input data that changes the prediction outcome. We put a specific emphasis on counterfactual explanations that are applicable to a wide range of data types (e.g., time series) and that are robust to time and retraining.
Overall, this research program will lead to a deeper understanding of what makes decision processes more efficient, fair, and transparent.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2022-05413
Name nom
Vidal, Thibaut
Department département
Mathematics and Industrial Engineering
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
2022
Award amount
52000
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
802
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Information systems and technology
Area of application fr
Systèmes d'information et technologie de l'information
Research subject code
1601
Research subject group en
Industrial engineering
Research subject group fr
Génie industriel
Research subject en
Operations research and management science
Research subject fr
Recherche opérationnelle et science de la gestion
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
Interpretable Decision Processes in Data-Driven Supply Chains
Keywords
combinatorial optimization; machine learning; operations research; supply chains; vehicle routing; production management; interpretable ai
Application summary
Supply chains are under constant transformation as a consequence of disruptive events (e.g., COVID-19), shifts in consumer behavior and new transportation paradigms (e.g., crowdsourced deliveries, autonomous vehicles). The greater availability of data means that more sophisticated decision and control processes can be deployed in practice. However, as these algorithms become more widespread and complex, new concerns arise concerning decision transparency and reliability. As of today, the successful application of machine learning and optimization algorithms rests on a delicate equilibrium: while the methods should be as sophisticated as required, they should also remain as simple and interpretable as possible to achieve fairness and accountability.
To address these important challenges, this research program focuses on the aspects of algorithmic simplicity, explainability, and robustness in data-driven supply chains and production systems. Our goal is to progress towards decision-support algorithms that are competitive in terms of economic performance, but that also possess the qualities mentioned above. To achieve these goals, we pursue cross-disciplinary methodological studies building upon machine learning techniques (e.g., decision tree-based predictors) and combinatorial optimization algorithms (e.g., metaheuristics and mathematical programming algorithms).
� We start our research by integrating explainable machine learning algorithms in transportation and production optimization. We use the capabilities of machine learning algorithms in contexts where it is difficult to estimate the impact of the decisions efficiently, either due to sources of uncertainty (stochastic and robust optimization) or long planning horizons (strategic optimization). We aim to learn decision rules to 1) prevent and react to disruptions and 2) to learn the economic impact of strategic decisions, and embed these components in state-of-the-art optimization algorithms.
� In the second part of this program, we refine available machine learning models and algorithms to adequate them to our applications and improve their performance, using combinatorial optimization techniques. We will pursue our research on model compression (i.e., simplification of a learned machine learning model) for decision-tree and neural-network classifiers. Moreover, we use combinatorial optimization algorithms to produce trustworthy "counterfactual explanations," i.e., smallest perturbations of an input data that changes the prediction outcome. We put a specific emphasis on counterfactual explanations that are applicable to a wide range of data types (e.g., time series) and that are robust to time and retraining.
Overall, this research program will lead to a deeper understanding of what makes decision processes more efficient, fair, and transparent.