Neuromorphic Neural Network Device Signal Decimation and Modulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The use of neuromorphic elements in neural networks for product-sum calculations leads to increased calculation time due to high input signal resolution, and applying voltage or current modulation results in resistance changes or breakdowns, necessitating precise pulse voltage control and high current detection resolution.
Innovation Solution
A neural network device with a decimation unit to convert input signals to lower step numbers, a modulation unit for pulse width or frequency modulation, and a weighting unit using neuromorphic elements to generate weighted signals, along with a current detection unit to manage the output current, thereby reducing calculation time and improving resolution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the resolution of the input signal to the neuromorphic element is increased, then the precision of signal processing is improved, but the calculation time increases
Solution Approach 1:
The patent divides the high-resolution input signal into multiple low-resolution segments that are processed separately and sequentially by the neuromorphic element. This segmentation allows the system to maintain high overall resolution while reducing the calculation time per segment, thereby resolving the contradiction between precision and speed.
Solution Approach 2:
The patent employs periodic pulse signals to modulate the input signal before feeding it to the neuromorphic element. By using periodic action with varying pulse widths or frequencies, the system can encode high-resolution information in a time-multiplexed manner, reducing the instantaneous calculation burden while maintaining precision.
2Ease of operation
If a high voltage is applied to the neuromorphic element to modulate the input signal, then the signal modulation capability is improved, but the resistance value changes or breakdown occurs
Solution Approach 1:
The patent changes the modulation parameter from voltage amplitude to pulse width or frequency. By using pulse width modulation (PWM) or pulse frequency modulation (PFM), the system achieves effective signal modulation without applying high voltages that would cause resistance changes or breakdown in the neuromorphic element, thus maintaining both operational capability and reliability.
3Measurement precision
If a high resolution current detection circuit is implemented, then the precision of output signal detection is improved, but the device complexity increases
Solution Approach 1:
The patent introduces an intermediary integration circuit that accumulates the output currents from multiple neuromorphic elements over time. This integration approach allows the use of simpler, lower-resolution detection circuits while maintaining high overall precision through temporal accumulation, thereby reducing device complexity without sacrificing detection accuracy.
Data Source
AI summary
A neural network device includes a decimation unit configured to convert a discrete value of an input signal to a discrete value having a smaller step number than a quantization step number of the input signal on the basis of a predetermined threshold value to generate a decimation signal a modulation unit configured to modulate a discrete value of the decimation signal generated by the decimation unit to generate a modulation signal indicating the discrete value of the decimation signal, and a weighting unit including a neuromorphic element configured to output a weighted signal obtained by weighting the modulation signal through multiplication of the modulation signal generated by the modulation unit by a weight according to a value of a variable characteristic.