Grant: Discovery Grants Program - Individual
McMaster University
$41,000.00
Geometric Deep Learning with Applications to Stochastic Finance
FY 2024-25
Summary
Deep learning has become an increasingly prevalent computational technology in mathematical finance and applied stochastic analysis, even if the search for a principled deep learning framework suitable to problems from this area is an actively researched open problem. Any meaningful and flexible deep learning framework must recognize that several problems in stochastic analysis are defined through the geometric structure they encode, and the preservation of that geometric structure is essential to generating meaningful predictions. Examples of such geometries range from infinite-dimensional and non-linear spaces of probability measures, such as the Wasserstein space from Robust finance, the space of signatures of a rough path from pathwise rough integration theory, to discrete spaces such as graphs capturing the systemic risk of inter-bank loans. This research program proposes to solve this open problem by leveraging geometric deep learning, a branch of machine learning designed to encode and preserve the geometric structures existing in learning problems. To do so, this research program proposes to extend the mathematical foundations of geometric deep learning to accommodate infinite-dimensional and non-smooth geometries.
However, many of the geometries central to mathematical finance and stochastic analysis are very different than those studied in mainstream geometric deep learning. This is because many of the important geometries in mathematical finance are typically infinite-dimensional and often "non-smooth." In contrast, geometries studied in mainstream geometric deep learning are either described in finite dimensions (e.g. finite graphs, finite grids, or compact Riemannian manifolds) or smooth (e.g. hyperbolic Riemannian manifolds or Riemannian manifolds of mixed curvature used to learn low-dimensional representations of large graphs).
Therefore, to solve the problem of how to construct a meaningful deep learning theory for finance and stochastic analysis, we must first extend the mathematical foundations of geometric deep learning to much more general "infinite-dimensional" metric spaces. Accordingly, the primary objective of this research program is to develop a universal approximation theory for geometric deep learning models between general metric spaces. The secondary objective of this research program is to specialize those principled geometric deep learning models to encode the relevant geometric structures in mathematical finance to propose a meaningful, trustworthy, and broadly applicable foundation for deep learning in mathematical finance and in stochastic analysis.
The successful completion of this research program will help Canadian financial institutions design trustworthy AI-powered decisions, consequentially reducing the model risk placed on Canada's economy. Academically, Canada is a leader in AI research, groups such as Mila and the Vector Institute, this program's results will add to Canada's momentum in AI research.
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-04482
Name nom
Kratsios, Anastasis
Department département
Mathematics
Organization id
27
Institution établissement
McMaster 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
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
2805
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Learning and inference theories
Research subject fr
Théories de l'apprentissage et l'inférence
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
Geometric Deep Learning with Applications to Stochastic Finance
Keywords
geometric deep learning; mathematical foundation of artificial intelligence; deep learning in mathematical finance; computational finance; applied optimal transport; geometric machine learning
Application summary
Deep learning has become an increasingly prevalent computational technology in mathematical finance and applied stochastic analysis, even if the search for a principled deep learning framework suitable to problems from this area is an actively researched open problem. Any meaningful and flexible deep learning framework must recognize that several problems in stochastic analysis are defined through the geometric structure they encode, and the preservation of that geometric structure is essential to generating meaningful predictions. Examples of such geometries range from infinite-dimensional and non-linear spaces of probability measures, such as the Wasserstein space from Robust finance, the space of signatures of a rough path from pathwise rough integration theory, to discrete spaces such as graphs capturing the systemic risk of inter-bank loans. This research program proposes to solve this open problem by leveraging geometric deep learning, a branch of machine learning designed to encode and preserve the geometric structures existing in learning problems. To do so, this research program proposes to extend the mathematical foundations of geometric deep learning to accommodate infinite-dimensional and non-smooth geometries.
However, many of the geometries central to mathematical finance and stochastic analysis are very different than those studied in mainstream geometric deep learning. This is because many of the important geometries in mathematical finance are typically infinite-dimensional and often "non-smooth." In contrast, geometries studied in mainstream geometric deep learning are either described in finite dimensions (e.g. finite graphs, finite grids, or compact Riemannian manifolds) or smooth (e.g. hyperbolic Riemannian manifolds or Riemannian manifolds of mixed curvature used to learn low-dimensional representations of large graphs).
Therefore, to solve the problem of how to construct a meaningful deep learning theory for finance and stochastic analysis, we must first extend the mathematical foundations of geometric deep learning to much more general "infinite-dimensional" metric spaces. Accordingly, the primary objective of this research program is to develop a universal approximation theory for geometric deep learning models between general metric spaces. The secondary objective of this research program is to specialize those principled geometric deep learning models to encode the relevant geometric structures in mathematical finance to propose a meaningful, trustworthy, and broadly applicable foundation for deep learning in mathematical finance and in stochastic analysis.
The successful completion of this research program will help Canadian financial institutions design trustworthy AI-powered decisions, consequentially reducing the model risk placed on Canada's economy. Academically, Canada is a leader in AI research, groups such as Mila and the Vector Institute, this program's results will add to Canada's momentum in AI research.