A memristor is the fourth fundamental passive circuit element, along with a resistor, capacitor, and inductor. Its name stands for ‘memory resistor,’ reflecting the main property of such a device: its resistance depends on the history of its input, retaining a memory of past states.
Memristive devices are inherently non-linear, with history-dependent dynamics and hysteresis loops in their input-output relations. Because of these properties, memristors are excellent computational building blocks for neuromorphic computing architectures.
In practical terms, however, non-linear operations incur significant computational, memory, and energy costs. Thus, practical attempts to implement a memristor must confront these challenges to achieve genuine applicability. In collaboration with the University of Vienna, QUBO Technology has pioneered a quantum-optical memristor – a device leveraging quantum linear optics for energy-efficient operations, complemented by measurement feedback to achieve non-linear and memory effects. This breakthrough has enabled us to integrate all essential elements for a practical memristor, operating at the speed of light. Our device is capable of functioning in both classical and quantum regimes. In the latter, single photons execute operations within the quantum state space, facilitating computational speed-ups beyond classical capabilities.
The Quantum Memristor technology platform opens various application areas, including quantum neuromorphic architectures for ultra-high performance light-based computing, which we pursue in QUBO BrainTech.
QUANTUM
MEMRISTOR
RECENT
Experimental photonics quantum memristor
Michele Spagnolo, Joshua Morris, Simone Piacentini, Michael Antesberger, Francesco Massa, Andrea Crespi, Francesco Ceccarelli, Roberto Osellame & Philip Walther
Nature Photonics 16, 318–323 (2022)
Memristive devices are a class of physical systems with history-dependent dynamics characterized by signature hysteresis loops in their input–output relations. In the past few decades, memristive devices have attracted enormous interest in electronics. This is because memristive dynamics is very pervasive in nanoscale devices, and has potentially groundbreaking applications ranging from energy-efficient memories to physical neural networks and neuromorphic computing platforms. Recently, the concept of a quantum memristor was introduced by a few proposals, all of which face limited technological practicality. Here we propose and experimentally demonstrate a novel quantum-optical memristor (based on integrated photonics) that acts on single-photon states. We fully characterize the memristive dynamics of our device and tomographically reconstruct its quantum output state. Finally, we propose a possible application of our device in the framework of quantum machine learning through a scheme of quantum reservoir computing, which we apply to classical and quantum learning tasks. Our simulations show promising results, and may break new ground towards the use of quantum memristors in quantum neuromorphic architectures.
For full article please refere to nature.com
PUBLICATIONS
Experimental quantum speed-up in reinforcement learning agents
V. Saggio, B.E. Asenbeck, A. Hamann, T. Strömberg, P. Schiansky, V. Dunjko, N. Friis, N.C. Harris, M. Hochberg, D. Englund, S. Wölk, H. J. Briegel, P. Walther
Nature 591, 229–233 (2021)
As the field of artificial intelligence advances, the demand for algorithms that can learn quickly and efficiently increases. An important paradigm within artificial intelligence is reinforcement learning1, where decision-making entities called agents interact with environments and learn by updating their behaviour on the basis of the obtained feedback. The crucial question for practical applications is how fast agents learn2. Although various studies have made use of quantum mechanics to speed up the agent’s decision-making process3,4, a reduction in learning time has not yet been demonstrated. Here we present a reinforcement learning experiment in which the learning process of an agent is sped up by using a quantum communication channel with the environment. We further show that combining this scenario with classical communication enables the evaluation of this improvement and allows optimal control of the learning progress. We implement this learning protocol on a compact and fully tunable integrated nanophotonic processor. The device interfaces with telecommunication-wavelength photons and features a fast active-feedback mechanism, demonstrating the agent’s systematic quantum advantage in a setup that could readily be integrated within future large-scale quantum communication networks.
For full article please refere to nature.com