Artificial neural networks modeled after real brains can perform cognitive tasks.
A new study has shown that artificial intelligence networks built on human brain connectivity can perform cognitive tasks efficiently.
Researchers used MRI data from a large Open Science repository to reconstruct a brain connectivity pattern and apply it to an artificial neuron. An ANN, a computing system that uses multiple inputs and output units, is similar to the human brain. The Neuro (Montreal Neurological Institute–Hospital) and Quebec Artificial Intelligence Institute taught the ANN how to complete a cognitive memory task. They also observed the ANN’s performance.
This unique approach is two-fold. Connectomics was a term used to describe brain connectivity. However, it did not consider how the brain performs computations or functions. Secondly, traditional ANNs are arbitrary and do not reflect real brain network organization. Researchers hoped to learn more about brain connectivity to create ANN architectures that would allow them to develop new design principles for artificial neural networks.
The researchers found that ANNs with human brain connectivity performed cognitive memory tasks more efficiently and flexibly than other benchmark architectures. Neuromorphic neural networks could use the same architecture to support multiple learning abilities in different contexts.
Bratislav Misic is a senior author at The Neuro and a researcher. “Neuroscience is a branch of AI that has recently diverged from neuroscience. Artificial networks can help us understand brain structure and brain function. Using empirical data to build neural networks will help us discover design principles for better AI. Both will enrich our brain knowledge and help inform one another.
The study, which was published in the journal Nature Machine Intelligence on August 9, 2021, was made possible by the Canada First Research Excellence Fund, which was awarded to McGill University for Healthy Brains, Healthy Lives initiative, Canada Research Chairs, Fonds du Recherche du Québec – Nature et Technologies, Natural Sciences and Engineering Research Council of Canada, Fonds du Recherche du Canada – Sante, Canadian Institute for Advanced Research, Canada Research Chairs, Fonds du Recherche du Quebec et Technologies, and Centre UNIQUE.
The Neuro
The Neuro – Montreal Neurological Institute Hospital – is a leading destination for brain research, advanced patient care, and brain research. The Neuro was founded in 1934 by Dr. Wilder Penfield, a renowned neurosurgeon. It has since grown to be Canada’s most considerable neuroscience research and clinical center, as well as one of the most important in the world. The Neuro is uniquely placed to make a difference in treating and understanding nervous system disorders through its seamless integration of patient care, research, and training. The Neuro established the Tanenbaum Open Science Institute in 2016. It is the first institution to embrace Open Science. McGill University’s Montreal Neurological Institute is a McGill University teaching and research institute. The Neuroscience Mission at McGill University Health Centre includes the Montreal Neurological Hospital.
