A new study out of McGill University details how artificial intelligence (AI) networks modeled on the human brain can perform cognitive tasks efficiently.
The research was published in the journal Nature Machine Intelligence on August 9.
Reconstructing Brain Connectivity Pattern
The team of researchers first examined MRI data from a large Open Science repository before reconstructing a brain connectivity pattern. This brain connectivity pattern was then applied to an artificial neural network (ANN), which is a computing system that operates in a similar manner to the biological brain.
The ANN was trained by a team of researchers at The Neuro (Montreal Neurological Institute-Hospital) and the Quebec Artificial Intelligence Institute, and it learned to perform a cognitive memory task. The team observed the ANN as it worked to complete the task.
There are two main factors that make this approach stand out when compared to previous ones on brain connectivity, or connectomics. Previous work often involved describing brain organization without paying much attention to how it performs computations and functions. On the other hand, traditional ANNs rely on arbitrary structures, which do not accurately represent the way real brain networks are organized.
With the new approach, the researchers integrated brain connectomics into the construction of ANN architectures, which they believed would provide insight into how the wiring of the brain supports specific cognitive skills. They were also looking to derive novel design principles for artificial networks.
The Team’s Findings
The team discovered that ANNs with human brain connectivity, or neuromorphic neural networks, performed cognitive memory tasks with more flexibility and efficiency than other architectures. The neuromorphic neural networks were capable of using the same underlying architecture for a wide range of learning capacities, which spread across various contexts.
Bratislav Misic is a researcher at The Neuro and the senior author of the research.
“The project unifies two vibrant and fast-paced scientific disciplines,” says Misic. “Neuroscience and AI share common roots, but have recently diverged. Using artificial networks will help us to understand how brain structure supports brain function. In turn, using empirical data to make neural networks will reveal design principles for building better AI. So, the two will help inform each other and enrich our understanding of the brain.”
The research was funded in part by Canada First Research Excellence Fund, which was awarded to McGill University for the Healthy Brains, Healthy Lives initiative. Funding was also provided by the Natural Sciences and Engineering Research Council of Canada, Fonds de Recherche du Québec — Santé, Canadian Institute for Advanced Research, Canada Research Chairs, Fonds de Recherche du Québec — Nature et Technologies, and Centre UNIQUE (Union of Neuroscience and Artificial Intelligence).