Neural Network: Boon for astronomers & neurobiologists, new discoveries on the horizon
The neural network became effective to solve astronomical problems such as three-body problems and form images from the human brain in neurology.
Innovative applications of neural networks have been found and astronomical problems are one of them. Researchers from the University of Cambridge have developed a neural network that can address the three-body problem. When three celestial bodies orbit each other, the gravitational pulls create an unpredictable system. The issue of determining the position of each mass in space and time had been there for astronomers.
Researchers claimed that they can solve the three-body problem faster with neural network than a conventional computer. This will give an advantage for astronomers to gain understanding of the behavior of star clusters that lead to the formation of black hole systems. The advanced software program, known as Brutus, has been used to generate nearly 9,900 simplified three-body scenarios. They have been fed to neural networks to teach how to solve. Then the program was applied to 5,000 unseen scenarios. Researchers found that AI solved the problems in very short duration, which was less than a second each. However, Brutus took nearly two minutes as AI determined the pattern instead of going step by step. Chris Foley, co-author and a biostatistician at the University of Cambridge, commented that if this neural network performs well, it will offer solutions in an unprecedented time frame and deeper questions will be answered.
Neural networks can give insights on human thoughts and the mind-reading machines will no longer be merely the work of science fiction. Researchers from the Russian corporation Neurobotics and the Moscow Institute of Physics and Technology developed a method to determine the brain activity of a person without using invasive brain implants. This work would lead to the development of non-invasive post-stroke rehabilitation devices that are monitored by brain signals.
It would lead to finding treatments for the cognitive disorder. For achieving such treatments and developing devices, the insights on how the information in the brain is encoded are essential and neurobiologists tried to understand it by studying the brain in real-time as a person watches a video. With the help of artificial neural networks and electroencephalography (EEG), a technique that records the brain waves through electrodes placed on the scalp is used to determine the thoughts of test subjects during the video. The unseen videos were shown to the subjects and EEG was recorded and applied to neural networks. the system gave images that can be categorized in nearly 90% of the cases.
Grigory Rashkov, co-author and a junior researcher at MIPT and a programmer at Neurobotics, outlined that though researchers did not expect EEG would contain sufficient information to form an image, reconstructing partial image was possible.
Neural network has become hugely popular and applicable for solving problems that were unsolvable before. However, they may not lead to various discoveries soon, but they lay the foundation for them. The industry for neural network is booming with a rise in applications. According to a portland based research firm, the neural network market is expected to reach $38.71 billion by 2023.
Article Contributor:
