Machine Learning Engine for Formal Verification Suitability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining suitability for formal verification in electronic designs are subjective, leading to misclassification and inefficient use of resources, as they rely on human interpretation of grading criteria rather than objective assessment.
Innovation Solution
A computer-implemented method using a machine learning engine, specifically a Bernoulli Naïve Bayes classification algorithm, to automatically determine whether an electronic design is amenable to formal verification by analyzing design features such as configuration complexity, design structures, and proof time requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If engineers use subjective criteria assessment to determine formal verification suitability, then the process allows flexibility in interpretation, but it leads to misclassification and inconsistent results
Solution Approach 1:
The patent replaces the manual, subjective assessment process with an automated machine learning classification system. The ML engine objectively evaluates design features and determines formal verification suitability without human intervention, eliminating the inconsistency and misclassification issues inherent in subjective engineering judgment while maintaining the flexibility to adapt to different design types through configurable features.
2Reliability
If engineers manually assess design suitability for formal verification, then they can apply expertise and judgment, but it consumes significant time and resources
Solution Approach 1:
The patent performs preliminary automated assessment of design suitability for formal verification before engineers invest significant time in detailed analysis. The machine learning engine quickly evaluates key design features and provides a preliminary classification, allowing engineers to focus their expert judgment only on cases that require further human review, thereby reducing overall assessment time while maintaining reliability.
Solution Approach 2:
The system enables designs to self-assess their suitability for formal verification through automated feature extraction and classification. The ML engine independently evaluates design characteristics without requiring manual engineering assessment, freeing engineers from routine classification tasks and allowing them to focus on more complex verification challenges.
3Manufacturing precision
If formal verification is applied to unsuitable designs, then comprehensive verification coverage may be achieved, but it wastes computational resources and time
Solution Approach 1:
The patent performs preliminary classification of design suitability for formal verification before the actual verification process begins. The machine learning engine evaluates design features and identifies unsuitable designs early, preventing waste of computational resources on designs that would not benefit from formal verification, while ensuring that suitable designs proceed to comprehensive verification coverage.
4Loss of energy
If formal verification is not applied to suitable designs, then resources are conserved, but valuable bugs may escape detection
Solution Approach 1:
The patent performs preliminary identification of designs suitable for formal verification using machine learning classification. This early identification ensures that designs capable of benefiting from formal verification are correctly flagged and proceed to comprehensive verification, preventing bug escape while avoiding unnecessary verification of unsuitable designs, thus optimizing resource conservation.
Data Source
AI summary
The present disclosure relates to a method for electronic design verification. Embodiments may include receiving, using a processor, an electronic design and providing at least a portion of the electronic design to a machine learning engine. Embodiments may further include automatically determining, based upon, at least in part, an output of the machine learning engine whether or not the at least a portion of the electronic design is amenable to formal verification.