Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

Université de Montréal

$33,000.00

An optimization framework for comparative, multi-species studies of visual systems

FY 2023-24

Summary
The visual system in our brain is responsible for processing optical input to form perceptions and guide actions. While this fundamental function of visual systems is preserved throughout evolution, comparisons across species have shown functional and anatomical differences in visual systems of different species. Current theories of vision do not take into account how evolution adapted visual systems to animals' natural environments. Given that evolution is the unifying idea of all biology, a multi-species comparative study can uncover both the general principles that govern the evolution and development of visual systems, and species' unique properties. Many theorists posit that visual systems are optimized to predict what happens next in animals' visual experience as they explore their natural environment. Informed by this idea, in this research program, we will test a predictive processing-based hypothesis: Because different species inhabit different environments, it follows that the visual systems adapted to predict these different visual experiences (through evolution or development) are likely to be different in anatomy and function; hence the cross-species differences of visual systems. Equipped with this hypothesis, we will examine the impact of visual experience on characteristics of visual systems and cross-species variations. Using simulation engines (such as Unity), we will generate video streams that mimic the visual experience of different species. We will use Artificial Neural Networks (ANN) for modeling visual systems. ANNs can capture parallel distributed computations (similar to the biological networks in the brain) and are simple enough to optimize. We will optimize ANNs to predict the simulated visual experience of different species. We will compare our simulated visual systems (species-specific ANNs) with biological visual systems of different species using neural data from mammalian visual systems (rodents, carnivores, and primates). Our work will advance our understanding of visual systems: how optical inputs are processed by brains to serve the visual life of animals in their natural habitat. In particular, we will obtain a better understanding of how ecological conditions of animals influence the characteristics of their nervous systems. In addition, our work will have a significant influence on the field of Artificial Intelligence (AI). The recent advances in AI have shown that scaling ANNs can lead to qualitatively different behaviors: larger ANNs are capable of more complex behaviors. Despite its effectiveness, scaling ANNs, in practice, is costly, energy inefficient, and can have negative effects on climate. The findings of our research program can shed light on scaling patterns across biological visual systems throughout evolution. These findings will inform scaling research in AI, and hopefully, create scaling algorithms that are more energy and cost efficient with less damaging climatic impact.
Program
Discovery Grants Program - Individual
Location
Quebec, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2023-03875
Name nom
Bakhtiari, Shahab
Department département
Psychology
Organization id
63
Institution établissement
Université de Montréal
Province en
Québec
Province fr
Québec
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
33000
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
1502
Committee name en
Biological Systems and Functions
Committee name fr
Systèmes et fonctions biologiques
Area of application code
405
Area of application group en
Environment
Area of application group fr
Environnement
Area of application en
Modelling and mathematical simulation of natural processes
Area of application fr
Modélisation et simulation mathématique des processus naturels
Research subject code
2800
Research subject group en
Artificial intelligence
Research subject group fr
Artificial intelligence (computer vision, use2603)
Research subject en
Intelligence artificielle
Research subject fr
Intelligence artificielle (vision artificielle, utiliser 2603)
Application title
An optimization framework for comparative, multi-species studies of visual systems
Keyword
Artificial neural networks; Computational modeling; Deep learning; Evolution; NeuroAI; Perception; Visual cortex; Visual learning; Visual systems
Application summary
The visual system in our brain is responsible for processing optical input to form perceptions and guide actions. While this fundamental function of visual systems is preserved throughout evolution, comparisons across species have shown functional and anatomical differences in visual systems of different species. Current theories of vision do not take into account how evolution adapted visual systems to animals' natural environments. Given that evolution is the unifying idea of all biology, a multi-species comparative study can uncover both the general principles that govern the evolution and development of visual systems, and species' unique properties. Many theorists posit that visual systems are optimized to predict what happens next in animals' visual experience as they explore their natural environment. Informed by this idea, in this research program, we will test a predictive processing-based hypothesis: Because different species inhabit different environments, it follows that the visual systems adapted to predict these different visual experiences (through evolution or development) are likely to be different in anatomy and function; hence the cross-species differences of visual systems. Equipped with this hypothesis, we will examine the impact of visual experience on characteristics of visual systems and cross-species variations. Using simulation engines (such as Unity), we will generate video streams that mimic the visual experience of different species. We will use Artificial Neural Networks (ANN) for modeling visual systems. ANNs can capture parallel distributed computations (similar to the biological networks in the brain) and are simple enough to optimize. We will optimize ANNs to predict the simulated visual experience of different species. We will compare our simulated visual systems (species-specific ANNs) with biological visual systems of different species using neural data from mammalian visual systems (rodents, carnivores, and primates). Our work will advance our understanding of visual systems: how optical inputs are processed by brains to serve the visual life of animals in their natural habitat. In particular, we will obtain a better understanding of how ecological conditions of animals influence the characteristics of their nervous systems. In addition, our work will have a significant influence on the field of Artificial Intelligence (AI). The recent advances in AI have shown that scaling ANNs can lead to qualitatively different behaviors: larger ANNs are capable of more complex behaviors. Despite its effectiveness, scaling ANNs, in practice, is costly, energy inefficient, and can have negative effects on climate. The findings of our research program can shed light on scaling patterns across biological visual systems throughout evolution. These findings will inform scaling research in AI, and hopefully, create scaling algorithms that are more energy and cost efficient with less damaging climatic impact.