PNAs: the bigger the brain, the smarter it is?

May 17, 2019 / BIOON / - a new study shows that increasing the size of neural circuits in the brain can improve learning performance, but this increased connectivity may also hinder learning Neuroscientists at the University of Cambridge predict that the "ideal" brain circuit has a size suitable for performing specific tasks The study, published in PNAS, looked at how neural circuits use extra connections to achieve faster, more accurate learning Photo source: PNAs research shows that adding obvious "redundant" neurons (the cells that make the brain work) and synaptic connections (making information flow from one neuron to another) to the network is actually a double-edged sword On the one hand, the increase of network connection can make the task easier to learn On the other hand, due to the inherent noise in the signal transmission connection, once the signal path exceeds a certain size, the increased connectivity will eventually hinder learning and task performance These findings reveal a new underlying reason why too many noisy connections can lead to learning disabilities, which are related to brain hyperconnections, including some developmental forms of autism Dr Timothy O'Leary, lecturer in Information Engineering and medical neuroscience who led the study, said: "our research shows that adding 'redundant' or redundant connections to brain circuits can actually improve learning efficiency "These extra connections - not absolutely necessary for brain function - can make a new task easier to learn "However, we find that if each new path adds' noise 'to its transmitted signal, the overall gain of learning performance will eventually be lost as the circuit size increases Therefore, we can predict the existence of a so-called "best point", an ideal brain circuit size for a specific task Although there is evidence that larger brains tend to exist in species with higher cognitive and learning abilities, the size of brain circuits may ultimately be limited by the need to learn efficiently at unreliable synapses In short, adding neurons and connections to the brain can help learning to some extent After that, the increase in the loop actually affects learning "Reference: dhruva venkita Raman et al Fundamental bounds on learning performance in natural circuits, Proceedings of the National Academy of Sciences (2019) Doi: 10.1073/pnas.1813416116