Revisiting neuroscientific hypotheses—new mechanisms of learning

The brain is a complex network of billions of neurons

For the past 70 years, a central assumption in neuroscience has been that the brain learns by adjusting the strength of synapses, based on the relative firing activity of the neurons that connect them

In an article published today in Scientific Reports, researchers from Bar-Ilan University in Israel reveal that the brain learns in a very different way than has been thought since the 20th century

"We have shown that efficient learning of the dendritic tree of individual neurons can artificially achieve near-uniform success rates for handwritten digit recognition

"This simplified learning mechanism represents a step toward the biological realization of the backpropagation algorithm, which is currently a core technology in artificial intelligence," added Ph.

Efficient learning of dendritic trees is based on experimental evidence from Kanter and his research team using subdendritic adaptation in neuronal cultures, as well as other anisotropic properties of neurons such as distinct spike waveforms, refractory periods, and maximum transmission rates

The brain's clock is a billion times slower than existing parallel graphics processors, but has comparable success rates in many perceptual tasks

New demonstrations of efficient learning of dendritic trees require new approaches to brain research and the generation of corresponding hardware aimed at implementing advanced artificial intelligence algorithms

Efficient dendritic learning as an alternative to synaptic plasticity hypothesis