Grant: Discovery Grants Program - Individual
Université de Sherbrooke
$41,000.00
Algorithms and Evaluation for Neuromorphic Multimodal Assistive Systems
EF 2024-25
Résumé
It is possible to train networks of Spiking Neurons (SN) with supervised algorithms that where initially designed for more conventional deep learning (DL) artificial neurons. SN are very interesting to the realization of efficient (in energy and speed) machine learning (ML) algorithms able to match results of DL. An `autonomous' device does not need external electrical power nor external computing resources.
i) Imagine an autonomous wearable device used to analyze the visual environment and represents it into sounds so that a blind person can get around the city and use public transportation. ii) Imagine another autonomous portable device which is used anywhere on earth to monitor health vital signs like coughs and provide scores on the potential for the patient to develop COVID-19 or pneumonia. iii) Finally imagine an autonomous Intelligent Agent device able to evolve and move in a real-life environment (f.e. mobile robot and/or drone) to collect plastic wastes on the seaside. For these 3 assistive application examples, we want a large electrical autonomy (many hours) while being able to run complex Machine Learning (ML) algorithms locally on energy-effective devices. Fortunately, SNN on dedicated platforms offer such potential for smart edge computing and wearable or mobile smart devices like the 3 ones introduced in this paragraph.
The long term goal of my research is on the development of autonomous intelligent applications viable when implemented with neuromorphic algorithms and to minimize their environmental footprint. In the next 5 years the short-term objectives will be:
1. to develop new supervised and unsupervised structural plasticity based learning rules for topology initialization of RSNN;
2. to develop conditioning and coupling between groups of spiking neurons for neuromorphic implementations of SNN;
3. to port on physical devices the 3 applications.
The evaluation of algorithms and applications will include 1) Energy consumption and speed of process- ing, 2) Size and type of required training data, 3) Classification and scores.
Design and development of algorithms inside this research program along with simulations, evaluations are planned during the first three years of the research (short-term) with 2 Ph.D. and Msc students. Then, the mid-term objective of porting to neuromorphic devices for: 1) COVID-19/tuberculosis detection and monitoring, 2) mobile agent for plastic waste collection, 3) vision to hearing for the blind is scheduled for the last 2 years (years 4 and 5) with 3 Msc.A. students.
Algorithms evaluations will be conducted with databases representative of the environment of the applications. This research program will contribute to create and maintain multidisciplinary HQP that will be highly viewed by AI Canadian companies.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Quebec, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2023-04847
Name nom
ROUAT, Jean
Department département
GEGI
Organization id
68
Institution établissement
Université de Sherbrooke
Province en
Québec
Province fr
Québec
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
306
Area of application group en
Energy resources
Area of application group fr
Ressources d'énergie
Area of application en
Energy efficiency
Area of application fr
Efficacité énergétique
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
Algorithms and Evaluation for Neuromorphic Multimodal Assistive Systems
Keywords
neuromorphic computing; signal to spike conversions; plasticity learning and initialization; audition & vision; neuromorphic algorithms evaluation; sensory substitution; multimodal features & representations; low energy smart devices; neural networks; machine learning
Application summary
It is possible to train networks of Spiking Neurons (SN) with supervised algorithms that where initially designed for more conventional deep learning (DL) artificial neurons. SN are very interesting to the realization of efficient (in energy and speed) machine learning (ML) algorithms able to match results of DL. An `autonomous' device does not need external electrical power nor external computing resources.
i) Imagine an autonomous wearable device used to analyze the visual environment and represents it into sounds so that a blind person can get around the city and use public transportation. ii) Imagine another autonomous portable device which is used anywhere on earth to monitor health vital signs like coughs and provide scores on the potential for the patient to develop COVID-19 or pneumonia. iii) Finally imagine an autonomous Intelligent Agent device able to evolve and move in a real-life environment (f.e. mobile robot and/or drone) to collect plastic wastes on the seaside. For these 3 assistive application examples, we want a large electrical autonomy (many hours) while being able to run complex Machine Learning (ML) algorithms locally on energy-effective devices. Fortunately, SNN on dedicated platforms offer such potential for smart edge computing and wearable or mobile smart devices like the 3 ones introduced in this paragraph.
The long term goal of my research is on the development of autonomous intelligent applications viable when implemented with neuromorphic algorithms and to minimize their environmental footprint. In the next 5 years the short-term objectives will be:
1. to develop new supervised and unsupervised structural plasticity based learning rules for topology initialization of RSNN;
2. to develop conditioning and coupling between groups of spiking neurons for neuromorphic implementations of SNN;
3. to port on physical devices the 3 applications.
The evaluation of algorithms and applications will include 1) Energy consumption and speed of process- ing, 2) Size and type of required training data, 3) Classification and scores.
Design and development of algorithms inside this research program along with simulations, evaluations are planned during the first three years of the research (short-term) with 2 Ph.D. and Msc students. Then, the mid-term objective of porting to neuromorphic devices for: 1) COVID-19/tuberculosis detection and monitoring, 2) mobile agent for plastic waste collection, 3) vision to hearing for the blind is scheduled for the last 2 years (years 4 and 5) with 3 Msc.A. students.
Algorithms evaluations will be conducted with databases representative of the environment of the applications. This research program will contribute to create and maintain multidisciplinary HQP that will be highly viewed by AI Canadian companies.