Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

University of Toronto

$71,000.00

Deep Learning Anywhere, Anytime: Enabling the Next Wave of Innovation through Hardware/Software Co-Design

FY 2024-25

Summary
Over the past few years, we have witnessed transformative advancements in computing. Now, our phones accurately comprehend voice commands, letting us seamlessly connect with loved ones and access information. Real-time language translation has dramatically improved, services such as ChatGPT assist with tasks ranging from information retrieval to coding, and numerous stories spotlight the role of computing in aiding individuals with disabilities. At the heart of these innovations lies Deep Learning-a collection of algorithms that enable computers to learn autonomously when provided with vast amounts of data. Though the foundational technology of Deep Learning has roots that stretch back decades, it's only in recent years-with the explosion of online data and the advent of advanced computing hardware-that we've observed its practical applications flourish. This has ignited a flurry of activity in both the industry and academia, leading to the creation of optimized hardware devices that drove innovation forward. However, as the pace of semiconductor technology decelerates, and such hardware designs have been optimized considerably, we face a potential stumbling block: our computing systems might not sustain further advancements in deep learning, stiflying future innovations. Regrettably, limited progress is anticipated from enhancements in semiconductor technology alone. The cornerstone for continued improvement and innovation is computer architecture which is concerned with creative ways of using the resources in a semiconductor-based design to ensure optimal execution speed and energy efficiency. Computer architecture is the defining factor that sets successful designs apart; especially when considering that there are only a few foundries remaining, and most designs utilize identical semiconductor technologies. Our five-year program is designed to bridge this gap by pioneering advancements in computer architecture and system software tailored for machine learning applications. We'll explore specialized hardware and delve into low-level software strategies that can supercharge computing capabilities, building on a solid foundation developed over the years. Our approach taps into the innate behaviors of applications-behaviors that might elude even the most skilled programmers and that emerge at runtime. Our techniques require no modifications to existing algorithms, yet they magnify optimization initiatives. Our designs automatically detect naturally occurring phenomena in machine learning applications and leverage them to enhance both performance and energy efficiency beyond the capacities of current methodologies. The result? Approaches that flawlessly merge into existing designs, amplifying their processing capability by tenfold across a myriad of devices and applications. This is indispensable for the ongoing evolution of deep learning applications and for perpetuating innovation in virtually all domains that harness machine learning.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2024-05979
Name nom
Moshovos, Andreas
Department département
Electrical and Computer Engineering
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
2024
Award amount
71000
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
1510
Committee name en
Electrical and Computer Engineering
Committee name fr
Génie électrique et informatique
Area of application code
513
Area of application group en
Manufacturing processes and products
Area of application group fr
Procédés et produits industriels
Area of application en
Electrical and electronic machinery and equipment (including computer hardware)
Area of application fr
Machinerie et équipement électriques et électroniques (y compris matériel informatique)
Research subject code
2701
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Computer hardware
Research subject fr
Matériel informatique
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
Deep Learning Anywhere, Anytime: Enabling the Next Wave of Innovation through Hardware/Software Co-Design
Keywords
energy-efficient computing hardware; high-performance computing hardware; deep learning hardware optimizations; deep learning software optimizations; deep learning applications; hardware/software assists
Application summary
Over the past few years, we have witnessed transformative advancements in computing. Now, our phones accurately comprehend voice commands, letting us seamlessly connect with loved ones and access information. Real-time language translation has dramatically improved, services such as ChatGPT assist with tasks ranging from information retrieval to coding, and numerous stories spotlight the role of computing in aiding individuals with disabilities. At the heart of these innovations lies Deep Learning-a collection of algorithms that enable computers to learn autonomously when provided with vast amounts of data. Though the foundational technology of Deep Learning has roots that stretch back decades, it's only in recent years-with the explosion of online data and the advent of advanced computing hardware-that we've observed its practical applications flourish. This has ignited a flurry of activity in both the industry and academia, leading to the creation of optimized hardware devices that drove innovation forward. However, as the pace of semiconductor technology decelerates, and such hardware designs have been optimized considerably, we face a potential stumbling block: our computing systems might not sustain further advancements in deep learning, stiflying future innovations. Regrettably, limited progress is anticipated from enhancements in semiconductor technology alone. The cornerstone for continued improvement and innovation is computer architecture which is concerned with creative ways of using the resources in a semiconductor-based design to ensure optimal execution speed and energy efficiency. Computer architecture is the defining factor that sets successful designs apart; especially when considering that there are only a few foundries remaining, and most designs utilize identical semiconductor technologies. Our five-year program is designed to bridge this gap by pioneering advancements in computer architecture and system software tailored for machine learning applications. We'll explore specialized hardware and delve into low-level software strategies that can supercharge computing capabilities, building on a solid foundation developed over the years. Our approach taps into the innate behaviors of applications-behaviors that might elude even the most skilled programmers and that emerge at runtime. Our techniques require no modifications to existing algorithms, yet they magnify optimization initiatives. Our designs automatically detect naturally occurring phenomena in machine learning applications and leverage them to enhance both performance and energy efficiency beyond the capacities of current methodologies. The result? Approaches that flawlessly merge into existing designs, amplifying their processing capability by tenfold across a myriad of devices and applications. This is indispensable for the ongoing evolution of deep learning applications and for perpetuating innovation in virtually all domains that harness machine learning.