- So far, we have known that the learning process in the brain is driven by synapses.
- Now researchers demonstrated that learning occurs in various long ramified arms, known as dendrites. .
- Dendritic learning occurs at quite faster rate than conventional synaptic learning.
The brain is the most complex part of the human body, containing over 86 billion neurons, and a less or more equal number of other cells. The interconnections of neurons make our brain active. The complete circuit is operated by a process called neurotransmission, in which nerve impulses trigger neurons to release neurotransmitters.
At least, that’s what we know since 1949, when Canadian psychologist, Donald Olding Hebb described how neurons contributes to psychological processes like learning. The neurons in the brain communicate with thousands of other neurons via links called synapses.
The pioneering work of Hebb indicates that neurons work as the computational modules, whereas learning occurs by altering synapses’ strength. However, scientists at the Bar-IIan University have demonstrated that learning occurs in various long ramified arms, known as dendrites, instead of synapses alone. It’s analogous to synapses’ slow learning process.
Newly Discovered Learning Process
According to the new study, dendritic learning occurs at quite faster rate than conventional synaptic learning. And, there are fewer adaptive factors per neurons, as compared to thousands of sensitive parameters in the synaptic learning. Scientists explained that it’s much more efficient for neurons to estimate their incoming signals close to other neurons.
Image credit: Wikimedia
For so many years, no one ever came up with any other approach. Since dendrites and synapses are connected in a series of neurons, it’s seemed insignificant to locate the exact region of the learning process.
The study also shows that weak synapses play a crucial role in the dynamics of the human brain. Instead of pushing learning factors to unreasonable fixed extremes — as proposed in synaptic learning — weak synapses produce oscillation of the learning factors.
The dendritic learning exhibits a self-controlled technique for obtaining oscillatory weight strengths. The emergence of slow and fast oscillations as an outcome of learning mechanism might be related to a source of transitory binding activities (among areas of macroscopic cortical), and high cognitive functions.
Scientists found that these oscillations were robust and had anisotropic nature of neurons. They had to be distinguished from oscillations coming from stochastic neuronal responses, which questions the synaptic slower learning rules.
Synaptic (left) and Dendritic learning (right) | Credit: Prof. Ido Kanter
The image explains it better: A neuron having 2 dendritic trees gathers incoming data through thousands of synapses (red valves). To put it in context, if neurons gathering their data are represented by a person’s fingers, the length of person’s hands would be comparable to skyscrapers.
While in the dendritic learning, these valves are placed at the close proximity to the neuron.
The events of learning occur in different regions of our brain, and therefore existing treatments of disordered brain functions need a serious re-evaluation.
Furthermore, learning processes are the basis of advanced machine/deep learning techniques and other AI methods. This research could open new possibilities for different forms of machine learning algorithms and AI based applications mimicking functions of the human brain, but with enhanced functionalities and faster processing speeds.