03:23 GMT27 November 2020
Listen Live
    Get short URL

    Neural networks, which can simulate our brain activity and learn, have graced humankind with a lot of perks, from navigating self-driving cars and synthesising human speech to images showing how we will age. However, the full potential of this is yet to be unlocked, as more sophisticated technologies could make better decisions than people.

    A research team from Tyumen State University in Russia has devised a new biomorphic neuron model and come up with concept principles for a new neural network by using it. The results of their research have been published in the journal Neural Computing and Applications.

    The Russian neuron model is similar to a real brain cell, both functionally and structurally. Like a nerve cell, the biomorphic copy created by the Siberian researchers consists of dendrites, soma, and an axon, which can be connected in many different ways and make its neural net more flexible.

    Additionally, the net, consisting of these neurons, can “think” and count faster. The creators focused on the medium frequency of electrical impulses, which innervate the cells, rather than on their form and, thus prolonging the time step.

    For biomorphic neuroprocessors the Russian scientists used memristors, which serve as a link between nerve cells, synapses, and allow for the development of an overbig biomorphic neural net, imitating the brain structure of our mind processing information – a cortical column.

    These artificial columns could be used to create a model of a brain part, called a neocortex, which will not need much computation power, making biomorphic neuronets faster and more energy-efficient than the existing count tools.


    Russian Scientists Improve Deep Learning Method for Neural Networks
    Russian Scientists Develop Neural Network for 3D Printing of Metal Products
    Psychologists: Neural Networks Will be Able to Predict One's Actions
    scientists, computing power, neurons, Tyumen, Russia
    Community standardsDiscussion