Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

Queen's University

$41,000.00

Interpretable and explainable learning with evolutionary computing

EF 2024-25

Résumé
Powerful artificial intelligence (AI) algorithms have provided solutions to numerous problems once thought impossible to solve. Human understanding of machine decision-making is deemed crucial for next-generation AI innovation. It is also a critical step toward equitable, inclusive, and ethical applications of AI. If we can understand a machine learning (ML) model, we will be able to better assess its fairness and trustworthiness on an application. This is particularly important for high-stakes tasks such as applications in medicine, job hiring, and criminal justice. Therefore, understanding AI systems is important both in its own right, and as a step toward evaluation fairness and ethical decision-making. The long-term goal of my research program is to develop learning methods that are transparent and understandable for critical applications. I propose to explore novel techniques based on evolutionary computing that learn interpretable AI solutions and provide more understandable explanations of individual predictions to better integrate AI into wider applications. Evolutionary computing, an approach to AI, employs a trial-and-error problem solving strategy and borrows ideas from how living organisms adapt through evolution. An evolutionary algorithm maintains a population of diverse candidate solutions whose quality, defined as fitness, can be assessed by comparison to the desired outcome. Fitter candidates will have the chance to be selected for stochastic changes and reproduction, and to survive to the next generation. Through multiple generations of selection, variation, and reproduction, the population adapts to the selection criterion and produces increasingly fitter solutions. Evolutionary algorithms have unique features including innovative learning, optimizing multiple objectives simultaneously, and providing diverse solutions that can be leveraged to derive models that are accurate, interpretable, and suitable for different purposes. The objectives of this proposed research include: 1) training of interpretable models using illumination evolutionary algorithms that make explicit the goal of maximizing the diversity of solutions; 2) design of post-hoc explanation techniques that translate and extract knowledge from trained models; and 3) development of visualization tools and interactive interfaces for expert users to interpret the learning results. The outcomes of this proposed research include algorithmic frameworks for interpretable and explainable learning and their applications to various domain-specific problems. It provides training for HQP who will develop algorithmic and theoretical research, heuristic algorithm application, and interdisciplinary biomedical data analysis skills. The proposed research program uniquely contributes to training AI experts who can provide tailored solutions to domain-specific problems and have equity, inclusion, and ethics as the foundation of their work.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Ontario, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
RGPIN-2023-03302
Name nom
Hu, Ting
Department département
Computing, School of - Computing, School of
Organization id
29
Institution établissement
Queen's University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
Award amount
41000
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
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1211
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Medical and health sciences
Area of application fr
Sciences médicales et de la santé
Research subject code
2713
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Algorithms
Research subject fr
Algorithmes
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 and explainable learning with evolutionary computing
Keywords
evolutionary computing; genetic programming; genetic algorithm; bio-inspired computing; bioinformatics
Application summary
Powerful artificial intelligence (AI) algorithms have provided solutions to numerous problems once thought impossible to solve. Human understanding of machine decision-making is deemed crucial for next-generation AI innovation. It is also a critical step toward equitable, inclusive, and ethical applications of AI. If we can understand a machine learning (ML) model, we will be able to better assess its fairness and trustworthiness on an application. This is particularly important for high-stakes tasks such as applications in medicine, job hiring, and criminal justice. Therefore, understanding AI systems is important both in its own right, and as a step toward evaluation fairness and ethical decision-making. The long-term goal of my research program is to develop learning methods that are transparent and understandable for critical applications. I propose to explore novel techniques based on evolutionary computing that learn interpretable AI solutions and provide more understandable explanations of individual predictions to better integrate AI into wider applications. Evolutionary computing, an approach to AI, employs a trial-and-error problem solving strategy and borrows ideas from how living organisms adapt through evolution. An evolutionary algorithm maintains a population of diverse candidate solutions whose quality, defined as fitness, can be assessed by comparison to the desired outcome. Fitter candidates will have the chance to be selected for stochastic changes and reproduction, and to survive to the next generation. Through multiple generations of selection, variation, and reproduction, the population adapts to the selection criterion and produces increasingly fitter solutions. Evolutionary algorithms have unique features including innovative learning, optimizing multiple objectives simultaneously, and providing diverse solutions that can be leveraged to derive models that are accurate, interpretable, and suitable for different purposes. The objectives of this proposed research include: 1) training of interpretable models using illumination evolutionary algorithms that make explicit the goal of maximizing the diversity of solutions; 2) design of post-hoc explanation techniques that translate and extract knowledge from trained models; and 3) development of visualization tools and interactive interfaces for expert users to interpret the learning results. The outcomes of this proposed research include algorithmic frameworks for interpretable and explainable learning and their applications to various domain-specific problems. It provides training for HQP who will develop algorithmic and theoretical research, heuristic algorithm application, and interdisciplinary biomedical data analysis skills. The proposed research program uniquely contributes to training AI experts who can provide tailored solutions to domain-specific problems and have equity, inclusion, and ethics as the foundation of their work.