18109710. MACHINE LEARNING SYSTEMS AND METHODS FOR CLASSIFICATION simplified abstract (Samsung Display Co., LTD.)
Contents
- 1 MACHINE LEARNING SYSTEMS AND METHODS FOR CLASSIFICATION
- 1.1 Organization Name
- 1.2 Inventor(s)
- 1.3 MACHINE LEARNING SYSTEMS AND METHODS FOR CLASSIFICATION - A simplified explanation of the abstract
- 1.4 Simplified Explanation
- 1.5 Potential Applications
- 1.6 Problems Solved
- 1.7 Benefits
- 1.8 Potential Commercial Applications
- 1.9 Possible Prior Art
- 1.10 Unanswered Questions
- 1.11 Original Abstract Submitted
MACHINE LEARNING SYSTEMS AND METHODS FOR CLASSIFICATION
Organization Name
Inventor(s)
Qisen Cheng of San Jose CA (US)
Kaushik Balakrishnan of San Jose CA (US)
Janghwan Lee of San Jose CA (US)
MACHINE LEARNING SYSTEMS AND METHODS FOR CLASSIFICATION - A simplified explanation of the abstract
This abstract first appeared for US patent application 18109710 titled 'MACHINE LEARNING SYSTEMS AND METHODS FOR CLASSIFICATION
Simplified Explanation
The patent application describes a classification system that calculates reference Shapley values for features of a data sample based on a first classification model, and then trains a second classification model using multi-task distillation to predict Shapley values for the features based on the reference values and a distillation loss, as well as predict a class label for the data sample based on the predicted Shapley values and a ground truth class label.
- The system includes one or more processors and memory with instructions for executing the described processes.
- The first classification model is used to calculate reference Shapley values for the features of a data sample.
- The second classification model is trained through multi-task distillation to predict Shapley values and class labels for the data sample.
Potential Applications
This technology could be applied in various fields such as finance, healthcare, and marketing for improving classification models and feature importance analysis.
Problems Solved
This technology helps in improving the interpretability and accuracy of classification models by incorporating Shapley values for feature importance.
Benefits
The system provides a more accurate understanding of feature importance in classification models, leading to better decision-making and model performance.
Potential Commercial Applications
Potential commercial applications include predictive analytics software, fraud detection systems, and personalized recommendation engines.
Possible Prior Art
Prior art may include research on Shapley values in machine learning models and multi-task learning techniques for model distillation.
Unanswered Questions
How does this technology compare to existing methods for feature importance analysis in classification models?
The article does not provide a direct comparison to existing methods for feature importance analysis, leaving the reader to wonder about the specific advantages of this approach.
What are the potential limitations or challenges in implementing this classification system in real-world applications?
The article does not address potential limitations or challenges that may arise when implementing this technology in practical settings, leaving room for further exploration of its feasibility and scalability.
Original Abstract Submitted
A classification system includes: one or more processors; and memory including instructions that, when executed by the one or more processors, cause the one or more processors to: calculate reference Shapley values for features of a data sample based on a first classification model; and train a second classification model though multi-task distillation to: predict Shapley values for the features of the data sample based on the reference Shapley values and a distillation loss; and predict a class label for the data sample based on the predicted Shapley values and a ground truth class label for the data sample.
