Rule-Set Evolution for Deep Learning Explainability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Deep learning models operate as black boxes, making it difficult to explain their functionality, especially in critical scenarios, and lack transparency, which can lead to costly failures and ethical concerns due to uninterpretable biases.
Innovation Solution
A system and method that uses rule-set evolution to generate transparent rule-set models equivalent to deep learning models, allowing for the evaluation and evolution of these models to provide explainability and replicate their performance in real-world problems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models are used to achieve high accuracy in classifications and predictions, then the model performance is improved, but the model becomes opaque and difficult to interpret
Solution Approach 1:
The patent creates a copy of the deep learning model's behavior through an evolved rule-set model. This rule-based copy replicates the predictive capabilities of the neural network while being interpretable through explicit if-then rules, thus resolving the contradiction between accuracy and interpretability
Solution Approach 2:
The patent introduces an intermediary evolutionary computation process that translates the black-box deep learning model into an explainable rule-set model. This intermediary serves as a bridge between the opaque neural network and human interpreters, maintaining predictive accuracy while enabling understanding
2Adaptability or versatility
If deep learning models are deployed in critical applications, then the capability to solve complex problems is improved, but the lack of transparency leads to ethical concerns and regulatory issues
Solution Approach 1:
By creating an interpretable copy of the deep learning model through rule-set evolution, the patent enables deployment in ethically sensitive domains. The rule-based representation allows stakeholders to understand, audit, and verify model decisions, ensuring ethical compliance while maintaining problem-solving capabilities
3Loss of information
If post-training interrogation methods are used to explain deep learning models, then some insights can be obtained, but the explanations are incomplete and open to interpretation
Solution Approach 1:
The patent replaces the mechanical process of post-hoc interrogation with an evolutionary computation process that proactively discovers accurate rules. Instead of querying the model after training, the system evolves rule-sets that precisely capture the model's decision logic, eliminating ambiguity and information loss
Data Source
AI summary
A system and method for adding explainability to deep learning models based on rule-set evolution is provided. A set of inputs from an input unit is received which comprises pre-generated deep learning models. Set of inputs is evaluated using pre-defined querying datasets. An output comprising outcomes of the evaluation is generated and mapped with each of the pre-defined querying datasets used for querying deep learning model. A new dataset is generated based on the mapping. Population of initial rule-set models is randomly generated based on a set of hyper parameters. The hyper parameters relate to configuration parameters used for generating population of initial rule-set models. An evolution process is carried out on generated rule-set models for evolving the rule-set models by using the generated new datasets. Lastly, the evolved rule-set model is executed to solve one or more real-world problems.