A group of scientists using automatic learning “rediscovered” the law of universal gravitation.
To do this, they formed a “network of graphic neurons” to simulate the dynamics of the sun, planets and large moons of the solar system of 30 years of observations. Then, they used symbolic regression to discover the analytical expression of the law of force implicitly learned by the neural network.
The result, as planned, was the law of Newton's universal gravitation.
Initially, the system had no idea of the masses of the sun, planets and physical constants. They were derived during research.
According to scientists, this study represents a key step towards carrying out the potential of automatic learning to accelerate scientific discovery.
More details can be found at https://arxiv.org/pdf/2202.02306.pdf
