This New Study Shows Artificial Neural Networks (ANN) Based On Human Brain Connectivity Can Perform Cognitive Tasks Efficiently

12036
Source: https://www.biorxiv.org/content/biorxiv/early/2020/11/11/2020.11.10.350876.full.pdf

Possibly a new breakthrough has been achieved in the domain of artificial intelligence. According to a new study, by a team of researchers from The Neuro (Montreal Neurological Institute-Hospital) and the Quebec Artificial Intelligence Institute, the artificial intelligence networks modelled on human brain connectivity are equipped to perform cognitive tasks efficiently and effectively. The study has been done via a sizable Open Science Repository by which the researchers tried to replicate and reconstruct the brain’s connectivity pattern. This was then applied to an artificial neural network (ANN) to achieve cognitive abilities like the human brain.

The Artificial Neural Network (ANN) 

The Artificial Neural Network (ANN) is a system that has both input and output units in abundance in similarity to the human brain. The researchers worked in two steps:

  1. Training the ANN to perform a cognitive memory task
  2. Observing how the stipulated task is completed

A Different Approach by the Researchers

This approach taken by the researchers is different from the traditional ones in the following ways:

  • The previous work done on brain connectivity and its application, namely connectomics, focused on describing the organisation of the brain. It did not look at the actual performance in computations and the functions. On the other hand, the new approach places a lot of focus on understanding the working and the performance for any given task.
  • The traditional ANNs used by most researchers before this did not reflect upon the arbitrary structures; by contrast, this new approach looks into the very construction of these structures to understand them wholly.

This novel approach by the researchers could pave the way for better understanding the brain’s wiring that primarily houses the cognitive skills and then subsequently improve the design principle for artificial intelligence and better its application in various fields. 

https://www.biorxiv.org/content/biorxiv/early/2020/11/11/2020.11.10.350876.full.pdf

The Findings of the Researchers

With this new approach to understanding the brain, one primary finding of the researchers was that the ANNs with human brain connectivity performed the cognitive tasks assigned to them much better. They were found to be more flexible and more efficient than all the other architectures that have been in place. It is being claimed that this new system made by the researchers used the same underlying architecture but, in turn, had broad learning capabilities across a vast spectrum.

This project undertaken is highly ambitious but, if successful, could establish new benchmarks across various industries of artificial intelligence and improve the very functioning of the same. Two of the most fast-paced disciplines Neuroscience and AI, have come together for this project. The data being collected and the working of the researchers could enrich both disciplines greatly.

Github: https://github.com/netneurolab/suarez_neuromorphicnetworks

Paper: https://www.biorxiv.org/content/biorxiv/early/2020/11/11/2020.11.10.350876.full.pdf