International business machines corporation (20240185027). MODEL TESTING USING TEST SAMPLE UNCERTAINTY simplified abstract

MODEL TESTING USING TEST SAMPLE UNCERTAINTY

Organization Name

international business machines corporation

Inventor(s)

Deepak Vijaykeerthy of Bangalore (IN)

Nishtha Madaan of Gurgaon (IN)

Swagatam Haldar of Kolkata (IN)

Aniya Aggarwal of NEW DELHI (IN)

Diptikalyan Saha of Bangalore (IN)

MODEL TESTING USING TEST SAMPLE UNCERTAINTY - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240185027 titled 'MODEL TESTING USING TEST SAMPLE UNCERTAINTY

Simplified Explanation

The abstract describes a method for training a proxy model to determine uncertainty scores corresponding to the output of a trained target model. These uncertainty scores are then used to select a subset of target model testing data for further analysis.

• The proxy model is trained using encoded representations of target model training data and corresponding labels.
• Uncertainty scores are computed for portions of target model testing data using the trained proxy model.
• A subset of target model testing data with uncertainty scores above a threshold is selected for further analysis using the trained target model.

Potential Applications

This technology could be applied in various fields such as finance, healthcare, and autonomous systems where uncertainty estimation is crucial for decision-making processes.

Problems Solved

This technology helps in identifying uncertain predictions made by the target model, allowing for more accurate and reliable decision-making based on the subset of data with higher uncertainty scores.

Benefits

The method provides a systematic approach to assess the uncertainty of predictions made by a target model, leading to improved decision-making and potentially reducing risks associated with incorrect predictions.

Potential Commercial Applications

• "Enhancing Decision-making with Uncertainty Estimation Technology"

Possible Prior Art

There may be prior art related to uncertainty estimation in machine learning models, but specific examples are not provided in the abstract.

How does this technology compare to existing methods for uncertainty estimation in machine learning models?

This article does not provide a comparison with other methods for uncertainty estimation, leaving the reader to wonder about the unique advantages of this approach.

What are the specific threshold values used for selecting the subset of target model testing data based on uncertainty scores?

The abstract does not mention the specific threshold values used for selecting the subset of data, leaving a gap in understanding the criteria for determining high uncertainty scores.

Original Abstract Submitted

using encoded representations of target model training data and a label corresponding to each portion of the target model training data, a proxy model to determine an uncertainty score corresponding to an output of a trained target model is trained. using the trained proxy model, a set of uncertainty scores is computed, each uncertainty score in the set of uncertainty scores corresponding to a portion of target model testing data in a set of target model testing data. a subset of the set of target model testing data is selected, the subset comprising a plurality of portions of target model testing data having an uncertainty score above a threshold uncertainty score. using the subset of the set of target model testing data, the trained target model.