Micro-Ring Resonator Array for Optical Neural Average Pooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current electronic computers face challenges in meeting high computing power and low power consumption demands, especially in big-data processing, and existing Optical Neural Networks are inadequate for simulating average pooling operations in neural networks.
Innovation Solution
A method involving a micro-ring-resonator array with unequal micro-ring radii and wavelengths, where electric current is applied to adjust transfer functions to achieve average pooling by summing optical signals outputted from the array through a photodiode.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Power
If conventional electronic computers are used for data processing, then computing power can be provided, but power consumption increases and transmission bottlenecks occur
Solution Approach 1:
The patent replaces the conventional electronic computing system with an optical computing system using micro-ring resonators. Optical signals substitute electrical signals, enabling computation through light-matter interactions in the micro-ring resonator array, thereby achieving high-speed processing with lower power consumption and eliminating electronic transmission bottlenecks
Solution Approach 2:
The patent utilizes wavelength as a key parameter to differentiate and process multiple optical signals simultaneously. By assigning different wavelengths to different data channels and using micro-ring resonators with specific resonance wavelengths, the system achieves parallel processing capability, dramatically increasing computing power while maintaining low power consumption
2Speed
If Optical Neural Networks are used for calculation, then high-speed concurrency and low power consumption are achieved, but the ability to simulate average pooling operations is insufficient
Solution Approach 1:
The patent segments the average pooling operation into multiple wavelength-specific processing channels. Each micro-ring resonator handles a specific wavelength (data channel), and through coordinated adjustment of their transfer functions, the system collectively implements the average pooling operation. This segmentation enables both high-speed optical processing and accurate simulation of neural network operations
Solution Approach 2:
The micro-ring resonator array is designed to perform multiple functions: wavelength filtering, signal modulation, and average pooling computation. By adjusting the transfer functions of individual resonators, the same hardware structure can adapt to different neural network operations, enhancing the versatility of optical neural networks while maintaining high-speed performance
3Device complexity
If micro-ring resonators with equal radii are used, then device structure is simple, but wavelength-specific resonance and transfer function adjustment capability is limited
Solution Approach 1:
The patent employs micro-ring resonators with asymmetric radii values. Each resonator has a uniquely sized ring structure that determines its resonance wavelength and transfer function characteristics. This asymmetric design enables each resonator to be independently tuned for specific wavelength processing, providing the adaptability needed for complex neural network operations like average pooling, while the modular array structure keeps overall device complexity manageable
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables high-speed, low-power consumption average pooling in neural networks, effectively addressing the limitations of existing technologies and providing an efficient solution for artificial intelligence calculations.
Implementation Method 1
by the to-be-treated optical signals with the unequal wavelengths, performing resonance with the corresponding micro-ring resonators
Implementation Method 2
feeding an optical signal outputted by the micro-ring-resonator array into a photodiode, to obtain an operation result of the average pooling of the neural network
Data Source
AI summary
The method includes: acquiring a plurality of to-be-treated optical signals with unequal wavelengths; inputting the to-be-treated optical signals into a micro-ring-resonator array, wherein the micro-ring-resonator array includes a plurality of micro-ring resonators that are connected in series; applying a corresponding electric current to the micro-ring-resonator array, to adjust a transfer function of each of the micro-ring resonators to reach a target value; and feeding an optical signal outputted by the micro-ring-resonator array into a photodiode, to obtain an operation result of the average pooling of the neural network.