Automatic phase compensation of a continuous-variable quantum-key-distribution system via deep learning

Phase compensation of a continuous-variable quantum key distribution via temporal convolutional neural network

“…In these figures, we find that the rate with TCN is closer to the rate without phase noise than that of the original pilot-based method (black lines), which indicates the role of TCN in reducing residual phase noise. Meanwhile, we find that a recent work uses an LSTM method for phase estimation [16] so we also show the results based on their method in figure 4(b) for comparison. We find that although the rates of these two schemes are close, TCN takes less time than LSTM.…”

“…However, in practice, there may exist uncertainty in the building models or the noise, which leads to the bad performance of the Kalman filter. Recently, Zhang et al propose a long short-term memory network (LSTM) model for phase compensation, which is not based on the mathematical model and noise statistics and reduces noise well [16]. Nevertheless, LSTM cannot perform parallel computation and has a long running time.…”

Optical Wireless Communications: Illuminating the Path for Cooper's Law

“…Such devicelevel advancement and development of multiplexing schemes suitable for the free-space QKD should be investigated to increase the secret key rate. Besides, the development of new protocols for diverse optical wireless channels [363], machine-learning-based environmental noise/interference suppression for QKD protocols [364], and machine-learningassisted precise and agile beam-steering/tracking techniques could be promising solutions to tackle the challenges. Methods for the system-level integration can include miniaturization through integrated silicon photonics, enabling high scalability [360] and hybridization with post-quantum cryptography [365] to facilitate the integration of QKD with modern wireless networks.…”

A Survey of Machine Learning Assisted Continuous-Variable Quantum Key Distribution

“…Generic CV-QKD protocol, where [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24] were applied during measurement, ref. [28] was applied to key sifting, ref.…”

“…While many traditional approaches have been suggested to overcome the limitations on CV-QKD, machine learning (ML) has recently been shown to have advantages in terms of phase error estimation and excess noise filtering [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20], state discrimination [21][22][23][24], parameter estimation and optimization [25][26][27], key sifting [28], reconciliation [29], and key rate estimation [30,31]. ML-based phase error estimation and noise filtering algorithms offer the potential of improved filtering capabilities due to their ability to map complex relationship between inputs and outputs based on the data alone, without being based on idealistic models that may not represent reality.…”