Achieving high-fidelity programmable photonic circuits (PPCs) is essential for photonic quantum computation and deep learning accelerators. Korean researchers at Seoul National University have reported heavy-tailed characteristics hardware pruning techniques for high-fidelity PPCs. The study appears in Nature Communications in April, 2023. 

A unitary operator is a building block for linear systems. Therefore, devising noise-robust, energy-efficient, and reconfigurable hardware for unitarities is essential for quantum information processing and deep learning accelerators. In a programmable photonic circuit, does every element contribute equally to the designed unitary operation? The answer to this question holds fundamental and practical importance in quantum physics and photonics for the development of more advanced hardware architecture. 

The researchers demonstrated heavy-tailed natures of programmable photonic circuits for universal unitary operations, revealing that some optoelectronic elements are more important than others. Even more surprisingly, they proved that it can be advantageous to remove insignificant, noisy elements, suggesting that “the bad is sometimes better to be removed”. This finding enables hardware pruning for general-purpose quantum circuits and photonic deep learning hardware, showing that in some cases, less can indeed be more for achieving optimal performance.

Quantum Pruning: Pruning insignificant elements leads to improved fidelity in realizing universal unitary operations, demonstrating that in some cases, less can indeed be more for achieving optimal performance.

[Reference] S. Yu & N. Park. “Heavy tails and pruning in programmable photonic circuits for universal unitaries.” Nature Communications 14, 1853 (2023). DOI: 10.1038/s41467-023-37611-9

[Main Author] Sunkyu Yu(Seoul National University), Namkyoo Park(Seoul National University)

* Contact: Prof. Sunkyu Yu (sunkyu.yu@snu.ac.kr), Prof. Namkyoo Park (nkpark@snu.ac.kr)