
Neural networks can mimic the human brain’s sleep cycles, according to a recent study from the University of California, San Diego, to combat catastrophic forgetting.
According to a news release from Maxim Bazhenov, Ph.D., a professor of medicine and sleep researcher at the University of California San Diego School of Medicine, “the brain is quite busy when we sleep, repeating what we have learnt throughout the day.” Sleep helps in memory reorganisation and delivers memories in their most effective form.
Sleep enhances rational memory, the ability to recall arbitrary or illogical associations between objects, people, or events, and protects against forgetting past memories, according to research by Bazhenov & colleagues.
They fail from time to time
Although artificial neurons function faster than the human brain, some times like a computer, it is clear that they require rest.
Contrarily, according to Bazhenov, “the human brain continuously learns and integrates new information into existing knowledge, and it normally learns best when new training is interrupted with sleep phases for memory consolidation.”
Bazhenov, a senior author, and colleagues discuss how biological models may reduce the risk of catastrophic forgetting in artificial neural networks, enhancing their utility in a variety of research areas.
Get some sleep not to forget
Researchers used spiking neural networks, which artificially mimic natural neural systems by transmitting information as discrete events (spikes) at specific times rather than continuously.
When the spiking networks were trained on a new task but with occasional off-line intervals that mirrored sleep, they noticed that catastrophic forgetting was decreased. The networks may replay earlier memories while sleeping, precisely like the human brain, according to the study’s authors, without specifically requiring prior training data.
It implied that these networks may continue to learn, just like humans or animals. It may be possible to improve memory in humans by better understanding how the brain processes information when we sleep. Better memory may result from improving sleep patterns.
Artificial neurons and the human brain
The human brain is made up of about 85 billion neurons, which weigh between 1250 to 1500 grammes on average. The brain is an organ of 2 hemispheres (cortex), and the two cortices of brain are separated from each other. These cortices’ morphological differences also affect how they perform. The right cortex of the brain functions like a parallel processor in comparison to a computer. On the other hand, the left cortex functions like a serial processor.
By examining human thought processes, artificial neuron studies typically try to create artificial instructions that are similar to these.
The study was published in PLOS Computational Biology