Artificial Neural Network Circuit Evaluation for Faster Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing complexity of integrated circuits and the time and resource-intensive nature of circuit verification processes necessitate a more efficient method for circuit evaluation.
Innovation Solution
Utilizing a trained artificial neural network model to generate a target circuit graph, vector, and evaluation data, incorporating neural networks such as CNN, RNN, GAN, and transformer models to predict circuit performance metrics like gain, power, and delay.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional circuit verification methods are used, then measurement precision is maintained, but evaluation time increases significantly
Solution Approach 1:
The patent creates a virtual copy of the circuit system through neural network models that replicate circuit behavior and characteristics. Instead of performing physical or detailed simulations on the actual circuit, the trained neural network model generates predicted performance data that mirrors real circuit responses, thereby maintaining accuracy while dramatically reducing evaluation time.
Solution Approach 2:
The patent performs preliminary training of neural network models using extensive circuit data and simulations before actual evaluation. This pre-training phase prepares the model to quickly predict circuit performance without requiring time-consuming simulations during the actual evaluation process, thus resolving the time-accuracy tradeoff.
2Reliability
If comprehensive circuit tests are performed, then reliability is improved, but computing resources are excessively consumed
Solution Approach 1:
The patent replaces resource-intensive physical simulations with a virtual neural network model that has been trained to replicate circuit behavior. This virtual copy maintains the reliability needed for verification while consuming minimal computing resources during the evaluation phase, as the heavy computational work was performed during the one-time training process.
Solution Approach 2:
The patent transforms the circuit evaluation problem from requiring detailed simulation of all circuit parameters to using a trained neural network model that has already processed and stored relationships between input parameters and output performance metrics. This parameter transformation allows quick evaluation without re-computing complex simulations.
3Manufacturing precision
If detailed feature extraction is performed, then manufacturing precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts only the most relevant circuit features and characteristics needed for performance prediction, rather than analyzing all possible circuit parameters. The neural network model is trained to identify and utilize key features that have the greatest impact on circuit performance, thereby maintaining manufacturing precision while reducing the effective complexity of the evaluation process.
Data Source
AI summary
A circuit evaluation method performed by a computing device according to an embodiment of the present disclosure. The method includes generating a target circuit graph including a plurality of nodes and edges based on target circuit data, generating a target circuit vector based on the generated target circuit graph, and generating circuit evaluation data based on the target circuit vector.