Australian scientists have created an analog computer in the form of a water slide, in which information is stored in a soliton - a surface wave that propagates in a non-linear medium and at the same time retains its shape and speed. This computer can be used to predict the chaotic behavior of complex systems. This is reported in an article published in the journal Europhysics Letters.

An analog computer is based on the so-called reservoir calculations, when the internal nonlinear dynamics of a medium is used for complex calculations. The biological brain can serve as such a reservoir, where the nonlinear dynamics of neural activity is used to predict the chaotic behavior of complex systems. Neurons form a network of random connections that transforms time-varying signals into a specific pattern (pattern) of activities.

A reservoir consisting of even a small number of neurons solves certain practical problems more efficiently than a computer program running on a supercomputer. The scientists suggested that a certain pattern depending on a time-varying signal (time series) can be represented as a soliton, and the reservoir in this case is water flowing down an inclined surface. However, there have been no experimental implementations of such an idea so far.

Related materials:On a new wave.Technologies are changing the world beyond recognition. What will it be like? November 12, 2021A place of attraction. How online commerce will make the lives of millions of people better and help save money December 7, 2021

In the new study, the scientists presented the time-varying signal as a series of ones and zeros. In this case, the unit corresponds to the changing pressure of water, which is spilled from the hose onto an inclined surface for 0.25 seconds, and zero corresponds to a constant pressure for 0.25 seconds. The chosen time intervals are optimal for soliton generation on the surface of flowing water.

A series of alternating zeros and ones generates solitons separated by equal intervals, and a signal in the form of a random series generates a complex profile of different waves, whose height does not always coincide with zeros and ones. This is because high amplitude waves run into low amplitude waves and absorb them. This principle underlies the memory of an analog computer and is similar to a system of neurons retaining traces of previous activity that affect the processing of new input data. To simulate the activity of 40 neurons, the soliton signal representing a random series was divided into small discrete steps.

To visualize the waves, a luminescent dye and a special camera were used. To test the reservoir computer, Mackey-Glass time series were used, which are characterized by non-linear chaotic behavior and are used to describe the cyclic operation of complex biological systems (for example, circadian rhythms). This time, the pressure of the water did not change in impulses, but continuously. The soliton profile obtained from the first half of the input dataset allowed scientists to predict with high accuracy each next step in the chaotic time series from the second half of the set using information from the previous steps.