Recurrent Neural Network Circuit Coefficient Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Reservoir computing systems face challenges in accurately and efficiently performing inference tasks due to the need for task-specific and data-type-dependent adjustments of reservoir characteristics, which current technologies struggle to address effectively.
Innovation Solution
An inference system comprising a recurrent neural network circuit and a control circuit that adjusts coefficients based on delay time periods, utilizing product-sum operation circuits with analog processing to optimize intermediate and output signal generation, enabling flexible and efficient pattern recognition and learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If reservoir characteristics are adjusted to be task-specific and data-type-dependent, then inference accuracy is improved, but system complexity and adjustment difficulty increase
Solution Approach 1:
The patent implements dynamic adjustment of reservoir characteristics through control circuits that modify connection weights and neuron parameters in real-time based on task requirements and data types. This allows the system to adapt its structure dynamically rather than requiring complex static configuration, resolving the contradiction between accuracy and complexity.
Solution Approach 2:
The system changes physical parameters of the reservoir circuit including connection weights, neuron activation thresholds, and time constants to optimize performance for different tasks. By adjusting these parameters dynamically through control circuits, the system achieves high inference accuracy without requiring fundamentally complex architectural changes.
2Measurement precision
If reservoir characteristics are adjusted to be task-specific and data-type-dependent, then inference accuracy is improved, but ease of operation deteriorates
Solution Approach 1:
The patent incorporates feedback mechanisms where control circuits monitor inference performance and automatically adjust reservoir characteristics accordingly. This closed-loop control simplifies operation by eliminating manual tuning requirements while maintaining high accuracy across different tasks and data types.
Solution Approach 2:
The system performs self-adjustment of its own characteristics through integrated control circuits that automatically optimize reservoir parameters based on incoming data characteristics and task requirements. This self-service capability improves ease of operation by making the system autonomous in its adaptation process.
3Productivity
If analog processing is used for product-sum operations, then processing speed is improved, but precision and stability deteriorate
Solution Approach 1:
The patent introduces digital control circuits as intermediary elements that manage the analog processing operations. These control circuits digitally determine optimal connection weights and parameters, then apply them to the analog reservoir circuit, combining the speed benefits of analog processing with the precision of digital control.
Solution Approach 2:
The system segments the processing into distinct functional blocks: digital control circuits for parameter determination and analog circuits for high-speed computation. This segmentation allows each part to operate in its optimal domain, maintaining both speed and precision/stability.
Data Source
AI summary
According to an embodiment, an inference system includes a recurrent neural network circuit, an inference neural network, and a control circuit. The recurrent neural network circuit receives M input signals and outputs N intermediate signals, where M is an integer of 2 or more and N is an integer of 2 or more. The inference neural network circuit receives the N intermediate signals and outputs L output signals, where L is an integer of 2 or more. The control circuit adjusts a plurality of coefficients that are set to the recurrent neural network circuit and adjusts a plurality of coefficients that are set to the inference neural network circuit. The control circuit adjusts the coefficients set to the recurrent neural network circuit according to a total delay time period from timing for applying the M input signals until timing for firing the L output signals.