Using active microparticles for artificial intelligence

- EN - DE
Schematic of the colloidal reservoir computer: polymer and gold-decorated partic
Schematic of the colloidal reservoir computer: polymer and gold-decorated particles that are controlled by a laser and perform calculations. Photo: Frank Cichos/University of Leipzig

Artificial intelligence with neural networks performs calculations digitally with the help of microelectronic chips. Physicists at Leipzig University have now realized a form of neural network that does not work with electricity but with so-called active colloidal particles. Their publication in the renowned journal "Nature Communications" deals with the use of these microparticles as a physical system for artificial intelligence and the prediction of time series.

"Our neural network belongs to the field of physical reservoir computing, in which the dynamics of physical processes such as water surfaces, bacteria or tentacle models of octopuses are used for calculation," explains Frank Cichos, whose working group developed the network with the support of ScaDS.AI. As one of five new AI centers in Germany, the research center with locations in Leipzig and Dresden has been funded since 2019 as part of the federal government’s AI strategy and supported by the Federal Ministry of Education and Research and the Free State of Saxony.

"In our realization, we use synthetic, self-propelled particles that are only a few micrometers in size," explains Cichos. "We show that these can be used for calculations and at the same time present a method that suppresses the influence of disturbing effects such as noise in the movement of the colloidal particles." Colloidal particles are particles that are finely dispersed in their dispersion medium (solid, gas or liquid).

For their experiments, the physicists have developed small units made of plastic and gold nanoparticles in which one particle rotates around another particle, driven by a laser. These units have certain physical properties that make them interesting for reservoir computing. Each of these units can process information, and many units form the so-called reservoir. We change the rotational movement of the particles in the reservoir using an input signal. The resulting rotation contains the result of a calculation," explains Dr. Xiangzun Wang. "To perform a certain calculation, the system needs to be trained like many neural networks."

The disruptive noise has been of particular interest to the working group. Since our system contains extremely small particles in water, the reservoir is subject to strong noise, similar to the noise that all molecules in a brain are subject to," reports Professor Cichos. "This noise, Brownian motion, severely disrupts the function of the reservoir computer and usually requires a very large reservoir to remedy. In our work, we have found that using past states of the reservoir can improve computer performance, allowing smaller reservoirs to be used for certain computations under noisy conditions."

This has not only made a contribution to the field of information processing with active matter, but has also developed a method that can optimize reservoir computing by means of noise reduction.

Publication in "Nature Communications":

"Harnessing synthetic active particles for physical reservoir computing"
DOI : 10.1038/s41467’024 -44856-5