Artificial Brains Learn To Adapt To Environment

guest author image

Lisa Winter

Guest Author

2095 Artificial Brains Learn To Adapt To Environment
Silvia Ferrari, Mechanical Engineering and Materials Science, Duke University

Connections between neurons don’t stay put. As we learn new things and respond to our environments over time, the brain’s electrical impulses and chemical signals change those connections, finding new routes between the memory associated with specific actions and stimulating the necessary motor function. Even this isn’t straightforward, as there are selective membranes and chemical gradients that affect the transmission of data. To say the very least, the brain is complex.

Artificial neural networks have tried to copy how the brain deals with memory, though they aren’t able to replicate some of the biological and chemical processes that interfere with synaptic activity. Silvia Ferrari of Duke University led a team of researchers that have created an artificial insect with a spiking neural network capable of learning from interacting with the environment. The research was presented last December at the IEEE 52nd Annual Conference on Decision and Control


"Although existing engineering systems are very effective at controlling dynamics, they are not yet capable of handling unpredicted damages and failures handled by biological brains," Ferrari said in a press release.

Spiking neural networks (SNN) create spikes of electricity to pass information along in the network, not completely unlike a natural synapse’s ability to alter the voltage of cells in order to transmit electrical signals. 

Ferrari’s team developed an algorithm that would allow an artificial insect running on an SNN to learn and navigate within an environment for a specific goal. Just as we learn from trial and error, the artificial insect did as well. After interacting with the environment, the SNN had to determine the importance of each action while working toward the goal. Over time, the network was able to draw on experience in order to make the best choices and accomplish a task, even when placed in an unfamiliar venue. 

"Our method has been tested by training a virtual insect to navigate in an unknown terrain and find foods," lead author Xu Zhang explained. "The nervous system was modeled by a large spiking neural network with unknown and random synaptic connections among those neurons.”


Ferrari’s team will build on this research and grow a living neural network comprised of lab-grown neurons. It is hoped that they will be able to replicate their success through exposure to light. This will help researchers better understand learned behavior through environmental sensory feedback.

These artificial brains could be used to control and navigate robotic equipment, as they could learn from their previous actions how to run more safely and efficiently. This technology could also have biomedical applications. In the future, it could be used to aid those suffering from Parkinson’s disease, epilepsy, Huntington’s disease, and other degenerative diseases of the central nervous system by regulating their loss of function.




Want to see a video of chemicals turning into memories in real time? Check it out here.


  • tag
  • memory,

  • learning,

  • artificial neural network,

  • spiking neural network