Micro-ring-resonator Hadamard Product Optical Neural Network
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Optical Neural Networks (ONN) solutions are inadequate for performing Hadamard product operations efficiently, as they primarily handle multiplication and addition based on convolution operations, while operations like those in Long Short-Term Memory (LSTM) artificial neural networks require extensive Hadamard product calculations, which are not effectively addressed by existing technologies.
Innovation Solution
A method utilizing a micro-ring-resonator component with wavelength division multiplexing to realize a Hadamard product by inputting optical signals with unequal wavelengths, where micro-ring-resonator groups with equal radii and controlled electric currents adjust effective refractive indexes and phases to achieve the desired output light intensity, simulating the Hadamard product operation in optical neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If exclusive devices are installed for convolution operations in Optical Neural Networks, then multiplication and addition operations can be performed, but the device cannot efficiently handle Hadamard product operations required by LSTM networks
Solution Approach 1:
The patent designs a unified optical computing device that can perform multiple operations including convolution, multiplication, addition, and Hadamard product using the same micro-ring-resonator-based architecture. The device achieves multi-functionality by configuring micro-ring-resonator groups with different coupling coefficients and applying different electric current patterns to the same physical structure, eliminating the need for separate exclusive devices for each operation type.
Solution Approach 2:
The patent employs dynamically adjustable coupling coefficients in the micro-ring-resonator groups by applying controllable electric currents. This dynamic configuration allows the same physical device to adapt its transfer functions for different computational operations (convolution, Hadamard product, etc.), enabling versatile operation without requiring multiple fixed-structure devices.
2Speed
If conventional optical computing methods are used, then processing speed is improved, but power consumption increases
Solution Approach 1:
The patent utilizes parameter changes in the micro-ring-resonator system, specifically adjusting the coupling coefficients and resonance conditions through controlled electric currents. This allows the device to perform complex computations like Hadamard products using optical interference effects rather than conventional electronic processing, achieving high-speed optical computation with lower power consumption by operating near the resonant conditions of the micro-ring structures.
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 effectively computes Hadamard products in optical neural networks, enhancing the capability to handle complex operations like those in LSTM networks with reduced power consumption and increased processing speed, aligning with future AI and high-data-volume requirements.
Implementation Method 1
applying a corresponding electric current to the micro-ring-resonator component, to obtain a result of the Hadamard product according to an outputted light intensity
Implementation Method 2
the micro-ring-resonator component comprises a plurality of micro-ring-resonator groups each of which is formed by two micro-ring resonators with equal radii
Data Source
AI summary
A method for realizing a Hadamard product, a device and a storage medium, includes: acquiring a plurality of to-be-treated optical signals with unequal wavelengths (S101); inputting the to-be-treated optical signals into a wavelength division multiplexer (S102); by using the wavelength division multiplexer, feeding the to-be-treated optical signals to a micro-ring-resonator component, wherein the micro-ring-resonator component includes a plurality of micro-ring-resonator groups each of which is formed by two micro-ring resonators with equal radii (S103); and applying a corresponding electric current to the micro-ring-resonator component, to obtain a result of the Hadamard product according to an outputted light intensity (S104). Accordingly, by using the micro-ring resonators as the base for realizing the solution of an artificial neural network, using the wavelength division multiplexer to feed the to-be-treated optical signals to the micro-ring-resonator component, and heating by using the electric current, the effective refractive indexes and the phases of the micro-ring resonators can be changed, and the result of the Hadamard product can be obtained according to the light intensity of the outputted optical signal, thereby realizing a simulated solution suitable for the Hadamard product in the optical neural network.