Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Grants Program - Individual

University of Toronto

$41,000.00

Type-safe machine learning programming

EF 2024-25

Résumé
Machine learning has achieved remarkable successes in numerous areas. However, these successes also pose many serious concerns due to the lack of reasonability and correctness guarantees, causing unpredictable failures. One source of such failures comes from implementation errors. Implementation errors are difficult to detect, since there can be many other potential reasons (such as noise in the training data) for such undesired behaviors, misleading programmers in their building and debugging process and making the resulting models hard to predict, let alone reason and understand. The correctness of model implementations relies largely on the language support involved in machine learning programming. However, state-of-the-art machine learning frameworks and libraries mostly build a tensor language interface in dynamically typed languages like Python with the core part written in high-performance low-level languages like C/C++. Those languages come with limited type systems, so certain errors can only be detected only at runtime. For example, programmers have suffered badly from shape errors caused by, e.g., multiplying two matrices with unmatchable dimensions. The long-term goal of the proposed research is to develop type-safe, efficient, and practical language design that builds more robust machine learning libraries and frameworks, so that we can more easily reason about machine learning programs and verify them against their specifications. In the next five years, the proposed research will focus on two main aims: (1) develop the semantics for tensors via array programming; and (2) develop toolchains for building and reasoning about machine learning programs. To achieve this goal, I will follow my recent developments in type systems, and applications of language semantics to machine learning languages and systems. The proposed research will advance the state-of-the-art machine learning programming techniques that greatly increase the expressiveness while striving to maintain the accessibility of type-safe machine learning programming, which is of fundamental importance for various machine learning domains. It will open many new avenues in the application of programming language theory to machine learning programming. The languages and frameworks developed will also lower the barrier to building reliable intelligent software for programmers with little formal semantics background.
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-2023-04632
Name nom
Xie, Ningning
Department département
Computer Science
Organization id
31
Institution établissement
University of Toronto
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
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2705
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Software and development
Research subject fr
Logiciels et développement
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
Type-safe machine learning programming
Keywords
programming language design; type theory; functional programming; array programming; dependent types; compiler; program verification; machine learning programming
Application summary
Machine learning has achieved remarkable successes in numerous areas. However, these successes also pose many serious concerns due to the lack of reasonability and correctness guarantees, causing unpredictable failures. One source of such failures comes from implementation errors. Implementation errors are difficult to detect, since there can be many other potential reasons (such as noise in the training data) for such undesired behaviors, misleading programmers in their building and debugging process and making the resulting models hard to predict, let alone reason and understand. The correctness of model implementations relies largely on the language support involved in machine learning programming. However, state-of-the-art machine learning frameworks and libraries mostly build a tensor language interface in dynamically typed languages like Python with the core part written in high-performance low-level languages like C/C++. Those languages come with limited type systems, so certain errors can only be detected only at runtime. For example, programmers have suffered badly from shape errors caused by, e.g., multiplying two matrices with unmatchable dimensions. The long-term goal of the proposed research is to develop type-safe, efficient, and practical language design that builds more robust machine learning libraries and frameworks, so that we can more easily reason about machine learning programs and verify them against their specifications. In the next five years, the proposed research will focus on two main aims: (1) develop the semantics for tensors via array programming; and (2) develop toolchains for building and reasoning about machine learning programs. To achieve this goal, I will follow my recent developments in type systems, and applications of language semantics to machine learning languages and systems. The proposed research will advance the state-of-the-art machine learning programming techniques that greatly increase the expressiveness while striving to maintain the accessibility of type-safe machine learning programming, which is of fundamental importance for various machine learning domains. It will open many new avenues in the application of programming language theory to machine learning programming. The languages and frameworks developed will also lower the barrier to building reliable intelligent software for programmers with little formal semantics background.