An Ensemble Deep Learning Model Based on Transformers for Long Sequence Time-Series Forecasting

Chu, Jie; Cao, Jingjing; Chen, Yujia

doi:10.1007/978-981-19-6135-9_21

Cited by 6 publications

(1 citation statement)

References 20 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, the transformer performed the best in modeling sequential data, becoming a popular model for ML. [25][26][27][28][29] Moreover, the transformer enables to capture global dependencies between feature and label by adopting self-attention, which has a higher parallelism and computational efficiency compared to recurrent neural network. 30 Based on these merits, it is powerful and promising to compensate the NLI for long-haul and large-capacity optical transmission system.…”

Section: Introductionmentioning

confidence: 99%

Low-complexity fiber nonlinear distortion mitigation for long-haul optical transmission based on transformer and triplets

Meng,

Cai,

Zhang

et al. 2024

Opt. Eng.

View full text Add to dashboard Cite

We propose and demonstrate an intrachannel fiber nonlinear impairments compensation scheme based on the transformer and simplified triplets. In comprehensive numerical experiments with a transmission distance of 1200 km for 30-GBaud dualpolarization 16-ary quadrature amplitude modulation signals, the Q factor gains of the proposed scheme are 2.48, 1.73, 0.98, and 0.31 dB compared to chromatic dispersion compensation, digital back propagation using 1-steps-per-span split-step Fourier method (1-StSp DBP), deep neural network (DNN), and 10-StSp DBP, respectively. The implementation complexity of the proposed scheme is reduced by 65% and 43% compared to 10-StSp DBP and DNN, respectively.

show abstract