Modeling complex organized quantum systems today is no easy task. The fact that the traditional methods are not suitable here, since with increasing system complexity the number of States increases exponentially. For example: system consisting of 100 quantum particles can be in any of the 1035 States. Even powerful supercomputers can not quickly deal with a miscalculation of the number of choices. And a group of scientists from the Swiss Federal Institute of technology presented a new method using neural networks that will speed up the process.
In my design, the experts used an innovative approach: instead of alternately calculate each possible state of the quantum system they are using neural network for its generalization. Scientists have developed a simplified version of a neural network and programmed it for modeling the wave functions of a quantum system. The resulting model can use an extensive set of numerical coefficients and one layer of “hidden” States. Based on these data, the system is able to calculate the system state based on a given set of conditions, bypassing the stage of calculation of each possible state.
To test their method, the researchers compared the data obtained using their algorithm, with the results that were obtained the “standard” way of calculating all possible options. As a result, when the same set of output data, the neural network coped with the task much faster. The developers hope that in the future such systems will help scientists conduct research more quickly and efficiently than it is now.