Quantum Memristors Combine AI and Quantum Computing

April 08, 2022 by Antonio Anzaldua Jr.

Memory capacity remains one of the enduring challenges to quantum computing. Now, Vienna researchers say they have found a way to scale quantum memristors.

Researchers at the University of Vienna in Austria believe they have discovered a new way to develop a quantum memristor that can be replicated for mass production. This discovery could represent an industry first: a quantum memristor that evolves beyond preliminary simulation demos and becomes an accessible product for superconducting circuitry. 

AI models often attempt to reproduce human reasoning

AI models often attempt to reproduce human reasoning, decision making, and environmental awareness. Image used courtesy of Equinox Graphics, University of Vienna


A classical memristor is a type of resistor that delivers resistance depending on the record of electrical signals that are applied to it. The memory of previous passing charges that have flowed through a memristor is stored physically in an internal state.

In the superconducting world, there are already quantized models of passive circuits such as inductors, capacitors, and resistors— but not memristors. 


What is a Quantum Memristor?

Physicists at the University of Vienna collaborated with the National Research Council (CNR) to demonstrate a new device, the photonic quantum memristor. This memristor includes a quantum processor and is able to operate on single photons.


Comparison of classical vs. quantum memristors

Comparison of classical vs. quantum memristors. Image (modified) used courtesy of Nature


A typical challenge for designers is finding discrete devices that can perform computations quickly and efficiently within neural networks. The "memory-resistor" or memristor could be a viable option for researchers to utilize in neural networks. The experiment that took place at the University of Vienna was led by Professor Philip Walther and Dr. Roberto Osellame, who presented a conceptual device that has the same behavior as the memristor while acting on quantum states, encoding and transmitting quantum information. 


Single Photons: The Key to Memristor's Success

One challenge to replicating this quantum memristor is the nature of memristors themselves; that is, the dynamics of a memristor tend to contradict typical quantum behavior. To address this barrier to mass production, the physicists behind this experiment used single photons, particles of light, which can propagate simultaneously along waveguides on a superposition of several paths. One path is used to measure the flux of photons that pass through the device. This path resides in a feedback scheme that modulates the transmission on the output end.


Diagram of an integrated photonics quantum memristor processor

Diagram of an integrated photonics quantum memristor processor. Image (modified) used courtesy of Nature


The group of researchers also ran simulations for these quantum memristors, demonstrating that optical networks could learn on both a classical and quantum task. The team concluded that there is only one limiting factor to scaling quantum memristors: the challenge of achieving a single-photon rate.

One possible solution, they said, is to integrate optical and discrete components within the same chip to help mitigate noise during periods of high frequency in photonic platforms. 


Quantum Memristors Shine a Light on Neuromorphic Compupting

The researchers posit that quantum memristors may one day be used to develop photonic neuromorphic computing systems. Neuromorphic computing entails brain-like capacity and efficiency. These neuromorphic machines would, in theory, be able to learn their surrounding environments instantly, deconstructing obstacles in order to make real-time decisions. 

Storing vast amounts of information has been one of the ongoing challenges of quantum computing, which is why a quantum memristor is highly sought after. It seems that photonic memristor hardware may be the key to overcoming this memory barrier, brinigng neuromorphic computing systems closer to reality.



Learn More About Memristor Research