Inference System Rule Importance Calculation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing inference systems, such as those using Probabilistic Soft Logic (PSL) and Markov Logic Networks (MLN), calculate posterior probabilities but fail to provide the basis or process leading to these results, making it difficult for users to understand and improve the inference system, especially in real-world decision-making scenarios where transparency and reliability are crucial.
Innovation Solution
An inference system that calculates the importance of each rule in a rule set by excluding rules one at a time and outputs the rule and its importance, allowing users to visualize and understand the inference process, thereby enabling better decision-making and system improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If an inference system outputs only the inference result without the basis or process, then the calculation efficiency is maintained, but the interpretability and user understanding of the inference process deteriorates
Solution Approach 1:
The patent segments the inference process into distinct components: the inference engine that calculates results and the explanation generator that produces human-readable justifications. By separating these functions, the system can maintain efficient inference operations while adding interpretability through dedicated explanation generation without significantly increasing overall system complexity
Solution Approach 2:
The patent introduces an intermediary explanation generation unit that translates the internal inference process and rule applications into human-understandable justifications. This intermediary component bridges the gap between the complex inference engine and the user, providing interpretability without requiring changes to the core inference mechanism
2Loss of information
If the inference system provides detailed basis and process information, then the interpretability improves, but the information processing load and computational cost increases
Solution Approach 1:
The patent applies partial action by generating explanations that include only the most relevant rules and inference steps rather than exhaustively documenting every computational detail. The explanation generation unit selectively presents information about rule applications and reasoning paths that are most important for user understanding, avoiding the computational overhead of complete process documentation
Solution Approach 2:
The patent changes the parameter of information representation from internal computational formats to human-readable formats. The explanation generation unit transforms complex inference data into simplified narratives that convey the essential reasoning logic without requiring users to process the full computational detail, thereby reducing the effective information processing load
3Measurement precision
If the system calculates importance for each rule by excluding rules one at a time, then the rule significance assessment accuracy improves, but the calculation time and computational resources increase
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing rule importance metrics during system initialization or rule set updates. Instead of performing exhaustive exclusion calculations during every inference operation, the system prepares importance assessments in advance, significantly reducing the computational time required during actual inference while maintaining accurate rule significance assessment
Solution Approach 2:
The patent implements periodic action by updating rule importance metrics at specific intervals or when the rule set changes, rather than continuously recalculating them for every inference. This periodic update approach maintains accurate importance information while avoiding the excessive computational burden of continuous recalculation
Data Source
AI summary
An inference method according to the present invention in an inference system inferring a probability that an ending state holds based on a starting state and a rule set, the method includes: when a rule set derived by excluding one rule from rules constituting a first rule set is set as a second rule set, a probability that the ending state holds based on the starting state and the first rule set is set as a first inference result, and a probability that the ending state holds based on the starting state and the second rule set is set as a second inference result, calculating an importance being an indicator indicating magnitude of a difference between the first inference result and the second inference result; and outputting the rule and the importance of the rule, being associated with each other for each of the excluded rule.