Reservoir Computing Apparatus Using Memristor Array
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing reservoir computing systems face high implementation costs and complexity due to the need for delay feedback loops and time-sharing multiplexing, which also increase power consumption and reduce memory and high-dimensional mapping abilities.
Innovation Solution
A reservoir computing apparatus with a signal input circuit, a reservoir circuit comprising mask and rotating neuron sub-circuits, and an output layer circuit that uses a memristor array to perform dimension raising, nonlinear operations, and recursive connections, eliminating the need for auxiliary circuits and optimizing operation efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If delay feedback loops and time-sharing multiplexing are used in reservoir computing systems, then the system can implement recursive connections and dimension raising, but the device complexity and implementation costs increase
Solution Approach 1:
The patent extracts the feedback loop function from traditional reservoir computing architecture and integrates it directly into the neural network layer structure. By removing the need for separate delay feedback loops and time-sharing multiplexing circuits, the invention achieves recursive connections through the natural recurrent structure of the neural network layers themselves, thereby reducing device complexity while maintaining adaptability
Solution Approach 2:
The patent merges the functions of multiple auxiliary circuits (delay feedback loops, time-sharing multiplexers) into a unified neural network layer structure. By combining the recursive connection functionality directly into the layer architecture, the invention eliminates the need for separate auxiliary components, reducing overall system complexity while preserving the essential reservoir computing capabilities
2Adaptability or versatility
If delay feedback loops and time-sharing multiplexing are used in reservoir computing systems, then the system can perform high-dimensional mapping, but the power consumption increases
Solution Approach 1:
The patent removes the power-consuming auxiliary circuits (delay feedback loops and time-sharing multiplexing) from the system architecture. By achieving high-dimensional mapping through the intrinsic properties of the neural network layers without requiring these additional active components, the invention significantly reduces power consumption while maintaining the ability to perform high-dimensional transformations of input signals
Solution Approach 2:
The neural network layers serve themselves by inherently providing the recursive connection and dimension raising functions that previously required separate auxiliary circuits. The layers utilize their own structure and parameters to achieve high-dimensional mapping without needing external power-intensive support systems, thereby reducing overall power consumption while maintaining functionality
3Adaptability or versatility
If delay feedback loops and time-sharing multiplexing are used in reservoir computing systems, then the system can implement reservoir computing functions, but the memory ability is reduced
Solution Approach 1:
The patent extracts the memory function from the auxiliary circuits and embeds it within the neural network layer structure itself. By eliminating delay feedback loops that consume memory resources and instead using the layered architecture to naturally preserve temporal information, the invention maintains reservoir computing functionality while increasing available memory capacity for other purposes
Data Source
AI summary
At least one embodiment of the present disclosure provides a reservoir computing apparatus and a data processing method. The reservoir computing apparatus includes: a signal input circuit, configured to receive an input signal; a reservoir circuit, including a plurality of reservoir sub-circuits, in which each reservoir sub-circuit includes a mask sub-circuit and a rotating neuron sub-circuit, the mask sub-circuit is configured to perform a first processing on the input signal with a first weight to obtain a first processing result, and the rotating neuron sub-circuit is configured to perform a second processing on the first processing result to obtain a second processing result; and an output layer circuit, configured to multiply a plurality of second processing results by a second weight matrix to obtain a third processing result. The reservoir computing apparatus optimizes operation efficiency and reduces implementation costs.