Artificial neural networks modeled on real brains can perform cognitive tasks

A caller survey shows that artificial quality networks based connected quality encephalon connectivity tin execute cognitive tasks efficiently.

By examining MRI information from a ample Open Science repository, researchers reconstructed a encephalon connectivity pattern, and applied it to an artificial neural network (ANN). An ANN is simply a computing strategy consisting of aggregate input and output units, overmuch similar the biologic brain. A squad of researchers from The Neuro (Montreal Neurological Institute-Hospital) and the Quebec Artificial Intelligence Institute trained the ANN to execute a cognitive representation task and observed however it worked to implicit the assignment.

This is simply a unsocial attack successful 2 ways. Previous enactment connected encephalon connectivity, besides known arsenic connectomics, focused connected describing encephalon organization, without looking astatine however it really performs computations and functions. Secondly, accepted ANNs person arbitrary structures that bash not bespeak however existent encephalon networks are organized. By integrating encephalon connectomics into the operation of ANN architectures, researchers hoped to some larn however the wiring of the encephalon supports circumstantial cognitive skills, and to deduce caller plan principles for artificial networks.

They recovered that ANNs with quality encephalon connectivity, known arsenic neuromorphic neural networks, performed cognitive representation tasks much flexibly and efficiently than different benchmark architectures. The neuromorphic neural networks were capable to usage the aforesaid underlying architecture to enactment a wide scope of learning capacities crossed aggregate contexts.

"The task unifies 2 vibrant and fast-paced technological disciplines," says Bratislav Misic, a researcher astatine The Neuro and the paper's elder author. "Neuroscience and AI stock communal roots, but person precocious diverged. Using artificial networks volition assistance america to recognize however encephalon operation supports encephalon function. In turn, utilizing empirical data to marque neural networks volition uncover plan principles for gathering amended AI. So, the 2 volition assistance pass each different and enrich our knowing of the brain."

This survey is published successful the diary Nature Machine Intelligence connected Aug. 9, 2021.

---

---