Reservoir Computing Processor for Time-Series Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current calculating devices using neural network models face limitations in achieving high computing power, particularly in performing complex calculations such as reservoir computing, which require efficient time-series processing and high reliability.
Innovation Solution
A calculating device is designed with a processor that updates variables using specific functions, including a first function that incorporates parts of a second variable set and a data set, and a second function that updates a second variable based on the first variable and data set, allowing for efficient time evolution calculations and improved computing power through parallel processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional neural network models are used for reservoir computing, then the calculating device can perform time-series processing, but the computing power and processing efficiency are insufficient
Solution Approach 1:
The patent divides the reservoir computing system into multiple independent neural networks, each responsible for processing specific time-series data segments. This segmentation allows parallel computation across multiple networks, significantly improving computing power and processing efficiency while maintaining manageable complexity through modular architecture.
Solution Approach 2:
The patent introduces a temporal dimension by processing time-series data through multiple time steps and creating time-evolving representations. This dimensional transformation enables the system to capture temporal patterns and dynamics, enhancing computing capability for complex time-series processing tasks.
2Reliability
If complex time evolution calculations are performed to induce nonlinear oscillator behavior, then the reliability of reservoir computing is improved, but the computational cost and processing time increase
Solution Approach 1:
The patent pre-computes and stores time-evolution trajectories and nonlinear oscillator characteristics before actual inference. This preliminary computation allows the system to quickly retrieve and apply pre-characterized time evolution patterns during processing, maintaining high reliability through accurate temporal modeling while significantly reducing real-time processing time.
Solution Approach 2:
The patent creates copies of time-evolution patterns and nonlinear oscillator behaviors through multiple neural networks that replicate and vary temporal dynamics. These copies can be independently processed and combined, ensuring reliable time-series processing while distributing computational workload to reduce processing time.
Data Source
AI summary
According to one embodiment, a calculating device includes a processor. The processor acquires a data set {s} and repeats a processing procedure. The processing procedure includes first and second variable updates. The first variable update includes updating an ith entry of a first variable xi by adding a first function to the ith entry of the first variable xi. The ith entry of the first variable xi is one of a first variable set {x}. A variable of the first function includes at least a part of a second variable set {y}. The second variable update includes updating an ith entry of a second variable yi by adding a second function and a third function to the ith entry of the second variable yi. The ith entry of the second variable yi is one of the second variable set {y}. The processor outputs at least a fourth function.