Canada Spends Logo

Explorez notre analyse du budget 2026

Grant: Discovery Launch Supplement

Simon Fraser University

$12,500.00

Theory-accelerated Deep Learning for the Human Genome

EF 2024-25

Résumé
This research will not involve Indigenous communities, societies or individuals, and therefore does not qualify as Indigenous research according to NSERC's definition. The nature of work in this proposal is to bridge the theoretical and practical aspects of deep learning, particularly in human genomics. Despite advancements in artificial intelligence (AI) and deep learning, a gap persists between theoretical understanding and practical applications, limiting fundamental comprehension and principled designs of deep learning methodologies. Theoretical work often focuses on simplified models and standard training settings, creating a void in understanding, especially in complex scientific domains, with challenges like low signal-to-noise ratios, label scarcity, and high dimensionality. This research scrutinizes crucial components of deep networks, focusing on data complexity, model architecture, and training dynamics, to optimize the AI pipeline through data compression, performance prediction, network architecture, and model scaling, in a zero-shot approach. This research program is important to both general and specialized deep learning users, for two reasons. First, our principles and empirical tools will target efficient designs and scale-up of foundation models in various scenarios, bridging the gap between theory and practical applications and providing robust, efficient, and scalable deep learning solutions. Second, our research will have profound implications for specialized fields, particularly scientific problems in human genomics, facilitating an efficient deep learning pipeline and accelerating scientific discovery. Anticipated outcomes of this proposal lie in five folds: 1. Establishing principled connections between training data complexity and deep network expressivity; 2. Optimizing deep network sizes and architectures based on data complexity and inductive bias. 3. Accelerating model pretraining/fine-tuning. 4. Developing cost-effective models and strategies on DNA sequences and related tasks; 5. Education, knowledge, and technology advancement in deep learning theory and applications, and interdisciplinary training of undergraduate and graduate students. Benefits of this proposal to the research field include addressing core questions in deep learning pipelines, enabling more refined scalable implementations and applications of neural network models in various domains, particularly computational biology. This proposal is expected to yield scientific contributions and enhance public visibility by sharing pretrained models on widely-used platforms. Benefits of this proposal to Canada encompass: 1) research: solidifying the position of Canadian academic institutions in global AI and machine learning; 2) technology: providing a catalyst for interdisciplinary collaborations, benefiting various Canadian enterprises and organizations; 3) education: furnishing excellent training opportunities for graduate students in computer science.
Programme
Discovery Launch Supplement
Emplacement
British Columbia, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data

Tel que publié

Application id
DGECR-2024-00476
Name nom
Chen, Wuyang
Department département
Computing Science
Organization id
5
Institution établissement
Simon Fraser University
Province en
British Columbia
Province fr
Colombie-Britannique
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
12500
Program id
DGECR
Program name en
Discovery Launch Supplement
Program name fr
Supplément Tremplin vers la découverte
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
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
2800
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Artificial intelligence (computer vision, use2603)
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
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
Theory-accelerated Deep Learning for the Human Genome
Keywords
artificial intelligence; deep learning; deep learning theory; machine learning; bioinformatics
Application summary
This research will not involve Indigenous communities, societies or individuals, and therefore does not qualify as Indigenous research according to NSERC's definition. The nature of work in this proposal is to bridge the theoretical and practical aspects of deep learning, particularly in human genomics. Despite advancements in artificial intelligence (AI) and deep learning, a gap persists between theoretical understanding and practical applications, limiting fundamental comprehension and principled designs of deep learning methodologies. Theoretical work often focuses on simplified models and standard training settings, creating a void in understanding, especially in complex scientific domains, with challenges like low signal-to-noise ratios, label scarcity, and high dimensionality. This research scrutinizes crucial components of deep networks, focusing on data complexity, model architecture, and training dynamics, to optimize the AI pipeline through data compression, performance prediction, network architecture, and model scaling, in a zero-shot approach. This research program is important to both general and specialized deep learning users, for two reasons. First, our principles and empirical tools will target efficient designs and scale-up of foundation models in various scenarios, bridging the gap between theory and practical applications and providing robust, efficient, and scalable deep learning solutions. Second, our research will have profound implications for specialized fields, particularly scientific problems in human genomics, facilitating an efficient deep learning pipeline and accelerating scientific discovery. Anticipated outcomes of this proposal lie in five folds: 1. Establishing principled connections between training data complexity and deep network expressivity; 2. Optimizing deep network sizes and architectures based on data complexity and inductive bias. 3. Accelerating model pretraining/fine-tuning. 4. Developing cost-effective models and strategies on DNA sequences and related tasks; 5. Education, knowledge, and technology advancement in deep learning theory and applications, and interdisciplinary training of undergraduate and graduate students. Benefits of this proposal to the research field include addressing core questions in deep learning pipelines, enabling more refined scalable implementations and applications of neural network models in various domains, particularly computational biology. This proposal is expected to yield scientific contributions and enhance public visibility by sharing pretrained models on widely-used platforms. Benefits of this proposal to Canada encompass: 1) research: solidifying the position of Canadian academic institutions in global AI and machine learning; 2) technology: providing a catalyst for interdisciplinary collaborations, benefiting various Canadian enterprises and organizations; 3) education: furnishing excellent training opportunities for graduate students in computer science.