Voltage Control Device for Neural Network Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge is to apply the limit operating voltage set during the learning process of a neural network to the inference process, as the PVT conditions during learning and inference may differ, leading to increased error rates in the inference neural network.
Innovation Solution
A voltage control device is developed, comprising a first neural network for learned information, a second neural network for unlearned information, an inference result determination unit, and a voltage determination unit. The voltage determination unit adjusts the voltage supplied to the neural networks based on the comparison between correct answer value data and inference result data, ensuring the voltage is optimized for accurate inference.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Use of energy by stationary object
If the limit operating voltage obtained during learning is applied to the inference neural network, then power consumption is reduced, but the error rate increases due to differing PVT conditions
Solution Approach 1:
The patent implements a feedback mechanism where the voltage control device monitors the actual error rate of the inference neural network and dynamically adjusts the supply voltage accordingly. When the error rate exceeds the acceptable threshold, the device increases the voltage; when the error rate is within the threshold, it maintains or reduces the voltage, thus resolving the contradiction between power consumption and reliability
Solution Approach 2:
The patent changes the voltage parameter dynamically based on real-time performance monitoring. Instead of using a fixed limit operating voltage, the system adjusts the voltage level according to the actual error rate conditions, allowing optimal power consumption while maintaining acceptable accuracy under varying PVT conditions
2Use of energy by stationary object
If the supply voltage is reduced to lower power consumption, then energy efficiency improves, but the error rate of the neural network increases
Solution Approach 1:
The patent introduces dynamic voltage adjustment capability to the neural network system. The voltage control device continuously monitors the error rate and dynamically modifies the supply voltage in real-time, allowing the system to operate at lower voltages when conditions permit while automatically increasing voltage when accuracy becomes compromised, thus resolving the static trade-off between power consumption and reliability
Data Source
AI summary
There is provided a voltage control device that automatically sets a limit operating voltage.Provided is a voltage control device including a first neural network, a second neural network, an inference result determination unit, and a voltage determination unit, in which the inference result determination unit has a function of comparing correct answer value data held by the inference result determination unit with inference result data of the first neural network to obtain determination result data, and the voltage determination unit has a function of outputting a voltage signal lower than a voltage supplied to the first neural network and the second neural network in a case where the correct answer value data and the inference result data match, and outputting a voltage signal higher than the voltage supplied to the first neural network and the second neural network in a case where the correct answer value data and the inference result data do not match, on the basis of the determination result data.