17937356. COUNTERFACTUAL SAMPLES FOR MAINTAINING CONSISTENCY BETWEEN MACHINE LEARNING MODELS simplified abstract (Capital One Services, LLC)
COUNTERFACTUAL SAMPLES FOR MAINTAINING CONSISTENCY BETWEEN MACHINE LEARNING MODELS
Organization Name
Inventor(s)
Samuel Sharpe of Cambridge MA (US)
Christopher Bayan Bruss of Washington DC (US)
Brian Barr of Schenectady NY (US)
COUNTERFACTUAL SAMPLES FOR MAINTAINING CONSISTENCY BETWEEN MACHINE LEARNING MODELS - A simplified explanation of the abstract
This abstract first appeared for US patent application 17937356 titled 'COUNTERFACTUAL SAMPLES FOR MAINTAINING CONSISTENCY BETWEEN MACHINE LEARNING MODELS
Simplified Explanation
The abstract of the patent application describes a computing system that uses machine learning and counterfactual samples to generate explanations for model predictions, determine text for explanations, and train models to generate consistent output. It also involves generating counterfactual samples based on user preferences by adjusting features indicated by the preferences.
- Explanation of the patent/innovation:
- Aggregating multiple counterfactual samples for generating machine learning explanations. - Using machine learning and counterfactual samples to determine text for model predictions. - Training machine learning models to produce consistent output. - Generating counterfactual samples based on user preferences by adjusting features.
Potential applications of this technology: - Enhancing transparency and interpretability of machine learning models. - Personalizing explanations for model predictions based on user preferences.
Problems solved by this technology: - Lack of transparency in machine learning models. - Difficulty in understanding and interpreting model predictions.
Benefits of this technology: - Improved trust in machine learning models. - Customized explanations for users based on preferences.
Potential commercial applications of this technology: - Explainable AI systems for healthcare, finance, and other industries. - Personalized recommendation systems based on user preferences.
Possible prior art: - Previous methods of generating counterfactual samples in machine learning models. - Techniques for training machine learning models to produce consistent output.
Unanswered questions: 1. How does the computing system handle conflicting user preferences when generating counterfactual samples? 2. What are the potential limitations or drawbacks of using counterfactual samples in machine learning explanations?
Original Abstract Submitted
In some aspects, a computing system may aggregating multiple counterfactual samples so that machine learning explanations can be generated for sub-populations. In addition, methods and systems described herein use machine learning and counterfactual samples to determine text to use in an explanation for a model's prediction. A computing system may also train machine learning models to not only determine whether a request to perform an action should be accepted, but also to generate output that is consistent with output generated by previous machine learning models. Further, a computing system may generate counterfactual samples based on user preferences. A computing system may obtain preferences and then apply a penalty or adjustment parameter such that when a counterfactual sample is created, the computing system is forced to change one or more features indicated by the preferences to create the counterfactual sample.
