Over-the-fiber Digital Predistortion Using Reinforcement Learning

Song, Jinxiang; He, Zonglong; Häger, Christian; Karlsson, Magnus; Amat, Alexandre Graell i; Wymeersch, Henk; Schröder, Jochen

doi:10.1109/ecoc52684.2021.9605972

Cited by 14 publications

(9 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The hardware impairments considered in this paper are restricted to the IQM nonlinearity and the limited ENOB of the DAC, while the PA is assumed to be linear, as the PA nonlinearity is negligible when compared to that of the MZM. However, it should be noted that the proposed approach can be readily applied to a more general setup where the other transmitter components are not idealized (e.g., nonlinear PA and bandwidth-limited DAC (see our previous work [50]).…”

Section: Numerical Resultsmentioning

confidence: 99%

“…In this paper however, we only focus on training of the PS filter, for which memory effects need to be taken into account. For the optimization of the mapper (i.e., NN1) or the DPD (i.e., NN3), we refer the reader to [23], [34] and our recent paper [50]. To that end, the parameters of the mapper, the DPD, and the demapper (i.e., NN4) are assumed to be pretrained and fixed during the PS filter training.…”

Section: Learning Without a Channel Modelmentioning

confidence: 99%

“…The reason for only learning the PS filter is that PS filter training contributes to most of the performance gain as it is shown in the ablation study. RL-based training of the mapper and the DPD can be found in [23], [34], and [50], respectively. Fig.…”

Section: Model-free Training Of the Pulse-shaping Filtermentioning

confidence: 99%

See 2 more Smart Citations

Model-Based End-to-End Learning for WDM Systems With Transceiver Hardware Impairments

Song

Häger

Schröder

et al. 2022

IEEE J. Select. Topics Quantum Electron.

Self Cite

View full text Add to dashboard Cite

We propose an autoencoder (AE)-based transceiver for a wavelength division multiplexing (WDM) system impaired by hardware imperfections. We design our AE following the architecture of conventional communication systems. This enables to initialize the AE-based transceiver to have similar performance to its conventional counterpart prior to training and improves the training convergence rate. We first train the AE in a singlechannel system, and show that it achieves performance improvements by putting energy outside the desired bandwidth, and therefore cannot be used for a WDM system. We then train the AE in a WDM setup. Simulation results show that the proposed AE significantly outperforms the conventional approach. More specifically, it increases the spectral efficiency of the considered system by reducing the guard band by 37% and 50% for a rootraised-cosine filter-based matched filter with 10% and 1% roll-off, respectively. An ablation study indicates that the performance gain can be ascribed to the optimization of the symbol mapper, the pulse-shaping filter, and the symbol demapper. Finally, we use reinforcement learning to learn the pulse-shaping filter assuming that the channel model is unknown. Simulation results show that the reinforcement-learning-based algorithm achieves similar performance as the standard supervised end-to-end learning approach assuming perfect channel knowledge.

show abstract

Section: Numerical Resultsmentioning

confidence: 99%

Section: Learning Without a Channel Modelmentioning

confidence: 99%

See 1 more Smart Citation

Model-Based End-to-End Learning for WDM Systems With Transceiver Hardware Impairments

Song

Häger

Schröder

et al. 2022

IEEE J. Select. Topics Quantum Electron.

Self Cite

View full text Add to dashboard Cite

show abstract

“…To alleviate the impact of spectrum deformation by pre-equalization at Tx and post-equalization at Rx, many studies based on ML have been reported [55][56][57][58][59][60][61]. The recurrent neural network (RNN), which is used for learning timedependent patterns or time-dispersive patterns including inter-symbol interference, can be widely used for pre-/postequalization because RNN structure is similar to that of conventional digital FIR filters.…”

Section: Digital Signal Processing Based On MLmentioning

confidence: 99%

Optical path management based on machine learning for optical networks

Shiraki

Mori

Hasegawa

2023

Next-Generation Optical Communication: Components, Sub-Systems, and Systems XII

View full text Add to dashboard Cite

The popularity of high-capacity communication services such as video streaming and cloud computing has accelerated the growth in IP traffic. In order to effectively manage and maintain networking systems, various intelligent technologies based on software-defined networking (SDN) have been widely studied. An SDN system that offers flexible optical path management exploiting optical performance monitoring, digital signal processing, and resource allocation is expected to realize higher capacity networks by lowering margins needed to offset system uncertainty. In this paper, we provide a comprehensive survey of optical path management schemes based on machine learning.

show abstract

“…Such a requirement brings special challenges for experimental systems where no gradient can be calculated. Nevertheless, methods such as gradient-free optimizers 49 or reinforcement learning 52 show great promise to address the challenge.…”

Section: End-to-end Learning Of Phase-noise Robust Communicationmentioning

confidence: 99%

Machine learning meets photonics

Da Ros,

Yankov,

Zibar

2023

Emerging Topics in Artificial Intelligence (ETAI) 2023

View full text Add to dashboard Cite

Machine learning (ML) is becoming a ubiquitous and powerful tool helping to address challenges in countless fields. Applications of ML addressing optics challenges have been extensively studied in recent years opening up new research directions. In particular, here, we review some of our current efforts and provide examples of successful applications of ML to the characterization of photonic devices, design, and modeling of optical subsystems, and complete end-to-end optical system optimization. ML and statistical tools can yield additional insight from measurement data, e.g. by targeted filtering of noise sources. They have also been shown to assist complex or inaccurate physics-based models through black and grey-box modeling of photonics components or subsystems. Such ML-aided models have enabled easier optimization and design (including inverse design) of optical systems.

show abstract

Over-the-fiber Digital Predistortion Using Reinforcement Learning

Cited by 14 publications

References 30 publications

Model-Based End-to-End Learning for WDM Systems With Transceiver Hardware Impairments

Model-Based End-to-End Learning for WDM Systems With Transceiver Hardware Impairments

Optical path management based on machine learning for optical networks

Machine learning meets photonics

Contact Info

Product

Resources

About