Scientists from the Faculty of Physics of the University of Warsaw and the Polish Academy of Sciences used photons to create a spiking neuron, the key element of the future photonic neural network processor. The so-called neuromorphic devices, i.e. systems that mimic the behavior of the biological brain, on which researchers are working, are the future of artificial intelligence, as they allow much faster and more efficient information processing. We can read about the results of their work in the latest “Laser and Photonics Review”.
The mammalian brain is one of the most complex and efficient systems in the world. Already in the 1990s, neurobiologists showed that a single area of macaque cortex was able to analyze and classify visual patterns in just 30 milliseconds, although each of the neurons involved in this process sends fewer than three messages in the form of electrical impulses. This is made possible by a large number of synapses – the connections between neurons – in the neural network of the macaque brain.
The human brain is a piece of even more powerful machinery. It consists of 100 billion neurons, each of which makes on average several thousand connections with other nerve cells. This creates a neural network of approx. 100 trillion connections, thanks to which our brain is able to recognize, reason and control movement at the same time – it performs trillions of operations per second, using only 20-25 watts of power. By comparison, conventional processors use ten times the power to recognize only a thousand different types of objects. This astonishing difference and extraordinary performance of the brain is due to, among other things, the biochemistry of neurons, the architecture of neural connections, and the biophysics of neural computing algorithms.
Society’s appetite for information is ever-increasing, so we need to process that information faster and more comprehensively. Conventional computing systems may not be able to meet the growing demand for more computing power while increasing energy efficiency. The solution to the problem may be the so-called neuromorphic devices that mimic the actions of the biological brain. They are the future of artificial intelligence, as they enable much faster and more efficient information processing in tasks such as image recognition.
Scientists from the Faculty of Physics of the University of Warsaw and the Polish Academy of Sciences in a paper published in “Laser and Photonics Review” proposed the use of photons in a way that allows the creation of cutting-edge neural networks. Krzysztof Tyszka from the Faculty of Physics of the University of Warsaw, who is the first author of the work, emphasizes that photonic systems ensure communication at the speed of light, low losses and low energy consumption. The advantage of photons is that their propagation takes place practically without loss of energy. – Unfortunately, because they interact in a relatively weak way, it is difficult to use them to perform computational operations in a manner analogous to electronic systems – adds the scientist.
– In our research, we propose a solution in which photons strongly interact with particles of very low mass, called excitons – explains Barbara Pietka from the Polariton Laboratory at the Faculty of Physics of the University of Warsaw. This strong interaction is possible when photons and excitons are trapped together in so-called optical microcavities, which force the repeated exchange of energy between them. This kind of synergy created in the microcavity between a photon and an exciton is so persistent that physicists call it a quasiparticle and refer to it as an excision polariton (or polariton for short).
Polaritons have unique properties, especially under the right conditions they can show a phase transition to a Bose-Einstein condensate. In such a situation, the previously independent multiple citizens become indistinguishable. – Based on our latest experiment, we were the first to observe that when polaritons are excited by laser pulses, they emit pulses of light in a manner that mimics the spiking of biological neurons – describes Magdalena Furman, Ph.D. student participating in research at the Polariton Laboratory at the Faculty of Physics of the University of Warsaw, This phenomenon is directly related to the Bose-Einstein condensation effect, which either inhibits or enhances the emission of pulses.
Andrzej Opala from the Institute of Physics of the Polish Academy of Sciences, who together with Michal Matuszewski developed the theoretical foundations that combine research on polaritons with the LIF model (Leaky Integrate-and-Fire model) of a neuron, adds that now the group is working to solve the problem of scalability, that is, the connection of many neurons in a network. – We propose to use a new computational paradigm based on pulse-coding information that activates a signal only when it reaches the neuron sequentially, at the right time – explains the researcher. Currently, neural networks use layers of interconnected neurons that fire pulses based on the importance assigned to each connection (in the mathematical description we refer to “weights”). In contrast to this type of solution, in the optical neural network developed by researchers from Poland, described in the journal Laser and Photonics Review, neurons are activated (ie become active) in response to a series of pulses, which may have different intensity and different time intervals. As with biological neurons stimulated by electrical impulses, there is a certain threshold above which this train of impulses reaching the neuron triggers a signal to be transmitted. Polaritons make it possible to mimic a biological system because only excitation with the right number of photons, above a certain threshold, leads to the formation of a Bose-Einstein condensate and then the emission of a small, picosecond-scale glow that is a signal for the next neuron.
Importantly, the sample, which was used by scientists to trap photons and observe the condensate of exciton polariton, was synthesized on site at the Faculty of Physics of the University of Warsaw, in the group of Wojciech Pacuski. The scientists arranged the atoms of different types of semiconductor crystals layer by layer through a molecular beam epitaxy to create a prototype photonic neuron. A temperature of 4K (liquid helium) was required to reach the Bose-Einstein condensate state. – Our further goal is to transfer the experiment from cryogenic conditions to room temperature – says Jacek Szczytko from the Faculty of Physics, University of Warsaw. – Research is needed for new materials that will allow obtaining Bose-Einstein condensates also at high temperatures. For photonic neurons to network, they must be able to transmit signals to each other. Ideally, the transmission direction, i.e. the wiring diagram, could easily be changed as needed.
– Scientists continue to face new challenges in their research into neuromorphic systems. Our new concept for recreating the cutting edge of biological neurons in the visual domain can be used to create a network and then a neuromorphic system in which information is sent orders of magnitude faster and in a more energy-efficient manner compared to existing solutions – concludes Krzysztof Tyszka.
An international team of scientists conducted research supported, among others, by the National Science Center (grants 2020/37B/ST3/01657, 2020/04/X/ST7 01379, 2020/36/T/ST3/00417), Center for Atomic Molecular and Optical Physics, and the European Union FET-Open Horizon 2020 program, grants ‘TopoLight’ (964770).
Physics and Astronomy at the University of Warsaw appeared in 1816 as part of the then Faculty of Philosophy. In 1825 the Astronomical Observatory was founded. Currently, the Faculty of Physics of the University of Warsaw consists of the following institutes: Experimental Physics, Theoretical Physics, Geophysics, Department of Mathematical Methods in Physics and Astronomical Observatory. The research covers almost all areas of modern physics, on scales from the quantum to the cosmological. The Faculty’s research and teaching staff consists of more than 200 academic professors, 81 of whom are professors. About 1,000 students and more than 170 PhD students study at the Faculty of Physics of the University of Warsaw.
K. Tyszka, M. Furman, R. Mirek, M. Krol, A. Opala, B. Seredynski, J. Suffczynski, W. Pacuski, M. Matuszewski, J. Szczytko, B. Pietka Leaky integration and fire mechanism in exciton-polariton condensates for a photonic spiking neuron
Laser & Photonics Reviews 2022, 2100660
Faculty of Physics, University of Warsaw
email: [email protected]
phone: +48 22 55 32 749
RELATED WWW SITES:
Polariton Group website
Website of the Faculty of Physics, University of Warsaw
Press Service of the Faculty of Physics, University of Warsaw
Optical microcavity as a pulsating neuron (visualization: Mateusz Krol, source: Faculty of Physics, University of Warsaw)
Laser & Photonics Review
Leaky integration and fire mechanism in exciton-polariton condensates for a photonic spiking neuron
Publication date of the article
#artificial #polar #neuron #step #photonic #system #mimics #human #brain #function