The chip contains 8,400 functional artificial neurons made of waveguide-coupled phase-change material. The neural network was trained to differentiate between German and English texts based on vowel frequency. © Jonas Schütte / AG Pernice
The chip contains 8,400 functional artificial neurons made of waveguide-coupled phase-change material. The neural network was trained to differentiate between German and English texts based on vowel frequency. © Jonas Schütte / AG Pernice
Tyler O'Neal, Staff Editor INTERCONNECTS

German researchers develop an adaptive optical neural network that connects thousands of artificial neurons

An international team of researchers has created a photonic processor that features adaptive neural connectivity. This processor is capable of processing data more efficiently and at a faster pace than traditional digital computers. Constructed from waveguide-coupled phase-change material, this processor comprises almost 8,400 functioning artificial neurons. It can differentiate between German and English texts based on vowel frequency. The German Research Association, the European Commission, and UK Research and Innovation have provided support for this research.

Supercomputer models for complex AI applications are pushing traditional digital computer processes to their limits. New types of computing architecture, which emulate the working principles of biological neural networks, hold the promise of faster, more energy-efficient data processing. A team of researchers has now developed an event-based architecture that uses photonic processors to transport and process data using light. Similar to the brain, this makes it possible to continuously adapt the connections within the neural network, which are the basis for learning processes.

The team of researchers used a network consisting of almost 8,400 optical neurons made of waveguide-coupled phase-change material. They showed that the connection between two neurons can indeed become stronger or weaker (synaptic plasticity) and that new connections can be formed, or existing ones eliminated (structural plasticity). In contrast to other similar studies, the synapses were not hardware elements but were coded as a result of the properties of the optical pulses.

Compared to traditional electronic processors, light-based processors offer a significantly higher bandwidth, making it possible to carry out complex computing tasks with lower energy consumption. The researchers aim to develop an optical computing architecture that will make it possible to compute AI applications in a rapid and energy-efficient way in the long term.

The non-volatile phase-change material can be switched between an amorphous structure and a crystalline structure with a highly ordered atomic lattice. This feature allows permanent data storage even without an energy supply. The researchers tested the performance of the neural network by using an evolutionary algorithm to train it to distinguish between German and English texts based on the number of vowels in the text.

The researchers received financial support from the German Research Association, the European Commission, and "UK Research and Innovation."

Swiss physicists build quantum network nodes with warm atoms

Tyler O'Neal, Staff Editor INTERCONNECTS

Communication networks need nodes at which information is processed or rerouted. Physicists at the University of Basel have now developed a network node for quantum communication networks that can store single photons in a vapor cell and pass them on later.

In quantum communication networks, information is transmitted by single particles of light (photons). At the nodes of such a network buffer elements are needed which can temporarily store, and later re-emit, the quantum information contained in the photons. A particle of light from the single photon source (below) is stored in the vapor cell (above). A simultaneously emitted second photon is revealed by a detector (right), which triggers the control laser pulse and thereby initiates the storage process. (Image: Department of Physics/University of Basel)

Researchers at the University of Basel in the group of Prof. Philipp Treutlein have now developed a quantum memory that is based on an atomic gas inside a glass cell. The atoms do not have to be specially cooled, which makes the memory easy to produce and versatile, even for satellite applications. Moreover, the researchers have realized a single photon source which allowed them to test the quality and storage time of the quantum memory. Their results were recently published in the scientific journal PRX Quantum.  

Warm atoms in vapor cells

“The suitability of warm atoms in vapor cells for quantum memories has been investigated for the past twenty years”, says Gianni Buser, who worked on the experiment as a Ph.D. student. “Usually, however, attenuated laser beams - and hence classical light - were used”. In classical light, the number of photons hitting the vapor cell in a certain period follows a statistical distribution; on average it is one photon, but sometimes it can be two, three, or none.

To test the quantum memory with “quantum light” – that is, always precisely one photon – Treutlein and his co-workers developed a dedicated single-photon source that emits exactly one photon at a time. The instant when that happens is heralded by a second photon, which is always sent out simultaneously with the first one. This allows the quantum memory to be activated at the right moment.

The single photon is then directed into the quantum memory where, with the help of a control laser beam, the photon causes more than a billion rubidium atoms to take on a so-called superposition state of two possible energy levels of the atoms.  The photon itself vanishes in the process, but the information contained in it is transformed into the superposition state of the atoms. A brief pulse of the control laser can then read out that information after a certain storage time and transform it back into a photon.

Reducing read-out noise

“Up to now, a critical point has been noise – additional light that is produced during the read-out and that can compromise the quality of the photon”, explains Roberto Mottola, another Ph.D. student in Treutlein’s lab. Using a few tricks, the physicists were able to reduce that noise sufficiently so that after storage times of several hundred nanoseconds the single-photon quality was still high.

“Those storage times are not very long, and we didn’t actually optimize them for this study”, Treutlein says, “but already now they are more than a hundred times longer than the duration of the stored single-photon pulse”. This means that the quantum memory developed by the Basel researchers can already be employed for interesting applications. For instance, it can synchronize randomly produced single photons, which can then be used in various quantum information applications.

South Korea demos a neuromodulation-inspired stashing system for the energy-efficient learning of a spiking neural network using a self-rectifying memristor array

Tyler O'Neal, Staff Editor INTERCONNECTS

Korea Advanced Institute of Science and Technology researchers have proposed a novel system inspired by the neuromodulation of the brain, referred to as a ‘stashing system,’ that requires less energy consumption. The research group led by Professor Kyung Min Kim from the Department of Materials Science and Engineering has developed a technology that can efficiently handle mathematical operations for artificial intelligence by imitating the continuous changes in the topology of the neural network according to the situation. The human brain changes its neural topology in real-time, learning to store or recall memories as needed. The research group presented a new artificial intelligence learning method that directly implements these neural coordination circuit configurations. A schematic illustrating the localized brain activity (a-c) and the configuration of the hardware and software hybrid neural network (d-e) using a self-rectifying memristor array (f-g).

Research on artificial intelligence is becoming very active, and the development of artificial intelligence-based electronic devices and product releases is accelerating, especially in the Fourth Industrial Revolution age. To implement artificial intelligence in electronic devices, customized hardware development should also be supported. However, most electronic devices for artificial intelligence require high power consumption and highly integrated memory arrays for large-scale tasks. It has been challenging to solve these power consumption and integration limitations, and efforts have been made to find out how the human brain solves problems.

To prove the efficiency of the developed technology, the research group created artificial neural network hardware equipped with a self-rectifying synaptic array and algorithm called a ‘stashing system’ that was developed to conduct artificial intelligence learning. As a result, it was able to reduce energy by 37% within the stashing system without any accuracy degradation. This result proves that emulating the neuromodulation in humans is possible.

Professor Kim said, "In this study, we implemented the learning method of the human brain with only a simple circuit composition and through this, we were able to reduce the energy needed by nearly 40 percent.”

This neuromodulation-inspired stashing system that mimics the brain’s neural activity is compatible with existing electronic devices and commercialized semiconductor hardware. It is expected to be used in the design of next-generation semiconductor chips for artificial intelligence.