The brain is one of the most complex parts of the body, and researchers and scientists to date haven’t been able to study all of its aspects in all their totality. However, it is not an old practice to compare the brain to a computer in both the disciplines of neuroscience and computer science. Our brains can perform almost all the tasks that we expect the computers to handle. Given this, it is clear that a better understanding of the brain would automatically help create better and more efficient computers. At the same time, if it is proved that computers are indeed like the human brains, then understanding the computation in the brains would consequently tell us when the machines would be able to match the standards. A productive flow would be established between the two fields making it mutually beneficial.
Working of the Brain and the Artificial Networks
Deep learning, per se, is a potent form of artificial intelligence, and it can be said that it has been based on the layered network of neurons (they are the cells that make up our brain, generally having three parts: dendrites, cell body, and axon). Each node that is found in the deep neural network is like an artificial neuron. And similar to the functioning of the neurons, the nodes also receive signals from the other nodes connected to them. Subsequently, mathematical tasks are then performed to gain output from the input. Furthermore, depending on the signal received, the node would also decide to send the signal to all the nodes available in the network, thereby largely sharpening and tuning a better-equipped algorithm. The brain’s functioning is also somewhat similar, although it is essential to note that these two processes are not carbon copies of each other.
In a new study, the team of scientists found that it took a five- to an eight-layer neural network, or nearly 1,000 artificial neurons, to mimic the behavior of a single biological neuron from the brain’s cortex. The researchers, however, clarify that the results are bound for some form of complexity and are not the exact measurement. Through this study, the researchers’ primary aim is to understand why the neurons are so complex. Once that is known, it could be a game-changer for designing much more capable neural networks and AI models.
To compare the brain and the computer, the main question that the researchers wanted to answer was as to how big of an artificial neural network would be required to match the behavior of a single neuron in the brain?
For the same purpose, numerous experiments were undertaken; firstly, 10,000 differential equations were used by the model to simulate precisely how and when the neuron would translate a series of input signals into a spike of its own. Then. Inputs were fed into the simulated neuron, and the outputs were recorded. Deep learning algorithms were then created on all the available data, and several layers kept increasing until the algorithm was 99 percent accurate at predicting the simulated neuron’s output given a set of inputs. The perfect production was found at least five layers but not more than eight, or around 1,000 artificial neurons per biological neuron. The algorithm wasn’t as complex as the original model but still could be categorized as complex.
The results were that the complexity of the neuron arose mainly due to the chemical receptors present in the dendrites. If the team has indeed developed an efficient algorithm, this result could prove beneficial for developing better and more efficient AI models. It is also possible that the results are not wholly accurate, given that the rat brain has been studied. More so, the computer and the brain could be different entities, but according to what research has shown, comparing and analyzing the two simultaneously has been fruitful.
Source Link: https://singularityhub.com/2021/09/12/new-study-finds-a-single-neuron-is-a-surprisingly-complex-little-computer/
The post Latest Research From Hebrew University Explains The Similarity Between Neurons and Artificial Neural Networks appeared first on MarkTechPost.