In the era of artificial intelligence and big data, there is a growing need for computing systems that can efficiently handle massively parallel tasks. Traditional computers, however, struggle with energy consumption and latency due to the physical separation between processors and memory. To address this, researchers have been exploring non-von Neumann architectures, such as in-memory computing, which performs computation directly within memory units.
In a new paper published in Light: Science & Applications , a research team led by Professor Xian-Min Jin from Shanghai Jiao Tong University and collaborators have demonstrated a quantum-enhanced in-memory stochastic computing system based on a room-temperature quantum memory. This system leverages the intrinsic randomness of quantum processes to perform computations securely and efficiently.
The core of the system is a quantum memory composed of cesium atoms, which generates correlated photons through controlled light-matter interactions. Computational tasks are encoded into the energy and timing of laser pulses that interact with the atomic ensemble. These pulses excite the atoms, leading to the probabilistic emission of Stokes and anti-Stokes photons. The photons are then detected and accumulated to yield computational results.
“Our system naturally supports stochastic computing operations such as addition and multiplication,” the scientists explained. “For example, addition is realized by accumulating Stokes photon counts, while multiplication is achieved by detecting coincidences between correlated Stokes and anti-Stokes photons.”
One of the key advantages of this approach is its security. Since the results are obtained through the accumulation of randomly generated photons, an eavesdropper intercepting a small portion of the data cannot discern meaningful information. This makes the system promising for secure remote computing.
Moreover, the team demonstrated that the use of quantum-correlated photons accelerates the computation process. Even with a low retrieval efficiency of 0.3%, the system outperforms classical stochastic computing methods in terms of coincidence rates and processing speed.
“This work shows that even imperfect quantum memories can be harnessed for practical computing tasks,” said Professor Jin. “We are optimistic that this will inspire new applications in quantum-enhanced computing and secure information processing.”
The research team believes that their room-temperature system could be further integrated with photonic chips and spatial multiplexing technologies to build compact and scalable quantum computing devices in the future.
Light Science & Applications
Quantum-enhanced Reconfigurable In-memory Stochastic Computing