一种基于光学储备池计算的信号补偿方法
By combining optical reservoir computation and a novel nonlinear mapping function of semiconductor optical amplifier, the problems of training complexity and high cost of traditional recurrent neural networks are solved, and efficient signal compensation effect is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- UNIV OF ELECTRONICS SCI & TECH OF CHINA
- Filing Date
- 2023-03-08
- Publication Date
- 2026-07-17
AI Technical Summary
Traditional recurrent neural networks suffer from problems such as complex training algorithms, large training volume, difficulty in determining network structure, and memory decay when dealing with time-related problems. Furthermore, the computational cost of SOA-based optical reservoirs is high and they are not easy to integrate.
A novel nonlinear mapping function based on semiconductor optical amplifiers and a reservoir calculation model for optical response are adopted. An echo state network is constructed by combining optical devices, and the network connection weights are trained using the ridge regression method. Signal compensation is performed through an optical reservoir.
It improved the signal-to-noise ratio by 8.32dB, reduced the bit error rate by 750 times, simplified the hardware implementation structure, and reduced energy consumption.
Smart Images
Figure CN116629332B_ABST