Grant: Discovery Grants Program - Individual
The University of British Columbia
$52,000.00
Learning Structured Models: Theory, Algorithms, and Quantum Oracles
EF 2024-25
Résumé
A principal preoccupation across the quantitative sciences is how to create mathematical models that fit real-world data. The models we construct from data allow us to make predictions based on past observations and to improve our conceptual understanding of the processes that generated the observed data. Because these models are constructed and deployed for so many important purposes throughout modern society --- from speech recognition to economic policy to medical imaging and diagnoses --- it is vitally important that these models be interpretable, reliable, and robust.
The long-term goal of my research is to develop mathematical theory, efficient algorithms, and robust software for data fitting that are portable across the whole spectrum of science and engineering fields. The core mission of my overall research program is to build accessible and user-friendly data-fitting tools with mathematical guarantees and flexible abstractions that work reliably for any type of dataset, and that can use a variety of computational resources with maximum efficiency. My recent research with collaborators and trainees in my research group has extended the mathematical foundations of optimization theory (algorithm complexity, convex duality, and model-parameter sensitivity analysis); developed probabilistic and deterministic recovery guarantees for a variety of data-fitting formulations; and built software used for a variety of scientific and engineering applications, including distributed learning, signal and image processing, reflection seismology, and signal reconstruction from incomplete and corrupted data.
My goal for the proposed research is to develop new theoretically supported algorithms that learn structured models from data. From a modeling perspective, the elements that constitute a structured model often represent key explanatory components that aid in model interpretability --- in other words, this is a type of generalized principle component analysis, but one that does not inherit the strict constraints of that particular approach. In addition to providing conceptual understanding, model interpretability is crucial for obtaining good generalizability and reliable inference in contexts where data is limited or expensive to obtain, such as in medical magnetic resonance imaging. From a computational perspective, structured models allow for significant algorithmic efficiencies, including reduced parameter models for faster training, reduced memory footprints, and improved convergence rates. I propose to build model-based deep learning methods built on structured optimization methods using well-defined oracles that are tractable for digital and quantum computers.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
British Columbia, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-05250
Name nom
Friedlander, Michael
Department département
Computer Science
Organization id
2
Institution établissement
The University of British Columbia
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2023
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
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1205
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Mathematical sciences
Area of application fr
Sciences mathématiques
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
Learning Structured Models: Theory, Algorithms, and Quantum Oracles
Keywords
mathematical optimization; large-scale convex optimization; structured optimization; statistical learning; mathematical programming; quantum computing
Application summary
A principal preoccupation across the quantitative sciences is how to create mathematical models that fit real-world data. The models we construct from data allow us to make predictions based on past observations and to improve our conceptual understanding of the processes that generated the observed data. Because these models are constructed and deployed for so many important purposes throughout modern society --- from speech recognition to economic policy to medical imaging and diagnoses --- it is vitally important that these models be interpretable, reliable, and robust.
The long-term goal of my research is to develop mathematical theory, efficient algorithms, and robust software for data fitting that are portable across the whole spectrum of science and engineering fields. The core mission of my overall research program is to build accessible and user-friendly data-fitting tools with mathematical guarantees and flexible abstractions that work reliably for any type of dataset, and that can use a variety of computational resources with maximum efficiency. My recent research with collaborators and trainees in my research group has extended the mathematical foundations of optimization theory (algorithm complexity, convex duality, and model-parameter sensitivity analysis); developed probabilistic and deterministic recovery guarantees for a variety of data-fitting formulations; and built software used for a variety of scientific and engineering applications, including distributed learning, signal and image processing, reflection seismology, and signal reconstruction from incomplete and corrupted data.
My goal for the proposed research is to develop new theoretically supported algorithms that learn structured models from data. From a modeling perspective, the elements that constitute a structured model often represent key explanatory components that aid in model interpretability --- in other words, this is a type of generalized principle component analysis, but one that does not inherit the strict constraints of that particular approach. In addition to providing conceptual understanding, model interpretability is crucial for obtaining good generalizability and reliable inference in contexts where data is limited or expensive to obtain, such as in medical magnetic resonance imaging. From a computational perspective, structured models allow for significant algorithmic efficiencies, including reduced parameter models for faster training, reduced memory footprints, and improved convergence rates. I propose to build model-based deep learning methods built on structured optimization methods using well-defined oracles that are tractable for digital and quantum computers.