Grant: Discovery Grants Program - Individual
University of Toronto
$59,000.00
The partitioning approach to learning: A signal processing perspective
FY 2024-25
Summary
Despite impressive progress in the last few years, learning in engineering systems is still domain-specific, application dependent, and subject to constraints. This program aims to introduce a signal-processing-inspired framework to address challenging learning problems. It advocates a partitioning approach to learning where the learner decomposes the task into a set of more straightforward, elementary learning tasks and then combines the elementary solutions to achieve the overall objective. Via partitioning, we can obtain a set of learning sub-tasks which are far easier to solve, smaller-scale, and more accessible to implement within given computational and communications constraints. Partitioning affords a wide variety of serial-parallel processing combinations of learning sub-tasks. Aggregating elementary learners in a simple, weighted average style can, most often, satisfies the overall learning objective. The framework provides deep insight into the learning problem's nature and the resulting learners' structure.
There exist learning problems of practical importance where a natural way to partition the learning problem is possible and where we can realize sub-learners of reduced complexity. This program will pursue three such learning problems as its short-term objectives: learn to estimate underlying probability densities in situations characterized by sequential dependencies in observations; learn to decide in partially observable dynamic environments modelled using continuous state and action spaces; learn to infer on distributed, server-less edge settings collaboratively. The program targets the advancement of the state-of-the-art and the development of practical learning designs, emphasizing implementations with reduced complexity and power consumption.
The view taken in this program is that of signal processing-based research. As several research communities contribute to learning, the program will not cover research questions that some may consider necessary. However, it unifies previously derived, seemingly non-connected results and introduces better theories and promising new solutions. It can open the door for innovative products in key technology areas for Canada, such as artificial intelligence (AI) and information communication technologies (ICT). Beyond that, the applicant hopes that the program will serve as inspiration for further studies in learning and spring new communities around specific interests in signal processing research. Lastly, but very importantly, the program will train scientists with deep knowledge of signal processing, machine learning, and statistical learning. Trainees will also develop professional skills in algorithmic development and system optimization and soft skills in terms of teamwork, collaboration, and industry interaction. Such skills are essential to lead research and innovation developments in Canada and beyond.
Program
Discovery Grants Program - Individual
Recipient
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data
As published
Application id
RGPIN-2023-04699
Name nom
Plataniotis, Konstantinos
Department département
ECE
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
59000
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
800
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 and communication services
Area of application fr
Services d'information et de télécommunications
Research subject code
2514
Research subject group en
Electrical and electronic engineering
Research subject group fr
Génie électrique et électronique
Research subject en
Digital signal processing
Research subject fr
Traitement des signaux numériques
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
The partitioning approach to learning: A signal processing perspective
Keywords
signal processing; machine learning; statistical machine learning; adaptive systems; dynamic programming; active inference; federated learning; reinforcement learning; edge learning; signal processing applications
Application summary
Despite impressive progress in the last few years, learning in engineering systems is still domain-specific, application dependent, and subject to constraints. This program aims to introduce a signal-processing-inspired framework to address challenging learning problems. It advocates a partitioning approach to learning where the learner decomposes the task into a set of more straightforward, elementary learning tasks and then combines the elementary solutions to achieve the overall objective. Via partitioning, we can obtain a set of learning sub-tasks which are far easier to solve, smaller-scale, and more accessible to implement within given computational and communications constraints. Partitioning affords a wide variety of serial-parallel processing combinations of learning sub-tasks. Aggregating elementary learners in a simple, weighted average style can, most often, satisfies the overall learning objective. The framework provides deep insight into the learning problem's nature and the resulting learners' structure.
There exist learning problems of practical importance where a natural way to partition the learning problem is possible and where we can realize sub-learners of reduced complexity. This program will pursue three such learning problems as its short-term objectives: learn to estimate underlying probability densities in situations characterized by sequential dependencies in observations; learn to decide in partially observable dynamic environments modelled using continuous state and action spaces; learn to infer on distributed, server-less edge settings collaboratively. The program targets the advancement of the state-of-the-art and the development of practical learning designs, emphasizing implementations with reduced complexity and power consumption.
The view taken in this program is that of signal processing-based research. As several research communities contribute to learning, the program will not cover research questions that some may consider necessary. However, it unifies previously derived, seemingly non-connected results and introduces better theories and promising new solutions. It can open the door for innovative products in key technology areas for Canada, such as artificial intelligence (AI) and information communication technologies (ICT). Beyond that, the applicant hopes that the program will serve as inspiration for further studies in learning and spring new communities around specific interests in signal processing research. Lastly, but very importantly, the program will train scientists with deep knowledge of signal processing, machine learning, and statistical learning. Trainees will also develop professional skills in algorithmic development and system optimization and soft skills in terms of teamwork, collaboration, and industry interaction. Such skills are essential to lead research and innovation developments in Canada and beyond.