Model Learning Device Using Grouped Size Comparison Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing model learning technologies face challenges in performing accurate learning using grouped uncoupled data, where input and output correspondences are not directly available, especially in scenarios involving sensitive information like annual income, and do not effectively handle repeated data collection over different periods and user groups.
Innovation Solution
A model learning device and method that acquire grouped uncoupled data and size comparison data, updating hyperparameters using a subgradient method and estimating optimization parameters using a gradient method to minimize specific objective functions, allowing for highly accurate model learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If uncoupled data is used for model learning to protect privacy, then privacy protection is improved, but model learning accuracy deteriorates
Solution Approach 1:
The patent introduces size comparison data as an intermediary to bridge the gap between uncoupled data and model learning. Instead of directly using uncoupled data alone, the patent combines it with size comparison data (which indicates whether one output is larger than another) to enable accurate model learning while maintaining privacy protection. This intermediary allows the system to infer relationships without requiring direct input-output correspondences.
Solution Approach 2:
The patent creates a composite learning approach by combining multiple types of data: uncoupled data (for privacy protection) and size comparison data (for maintaining learning accuracy). This composite data structure allows the model to learn from privacy-protected data while still achieving accurate predictions through the complementary information provided by size comparisons.
2Measurement precision
If size comparison data is added to uncoupled data for better learning accuracy, then model learning accuracy is improved, but data collection complexity increases
Solution Approach 1:
The patent changes the parameter of data format from traditional coupled (input-output pairs) or purely uncoupled data to a hybrid format that includes size comparison information. By transforming the data representation to include comparative relationships, the system achieves better learning accuracy without fundamentally changing the data collection process, thus avoiding increased complexity.
3Adaptability or versatility
If grouped uncoupled data is used from multiple periods and user groups, then adaptability is improved, but existing learning methods become inapplicable
Solution Approach 1:
The patent segments the learning process into distinct stages: first estimating parameters from uncoupled data, then refining them using size comparison data. This segmentation allows the system to handle grouped uncoupled data from multiple periods and user groups systematically, making existing learning methods applicable to this complex scenario through a structured two-step approach.
Data Source
AI summary
An aspect of the present invention acquires learning data including grouped uncoupled data acquired from a plurality of groups to be investigated, and grouped size comparison data. First, a processing of updating a hyperparameter using a first optimization method is executed on the obtained grouped uncoupled data, and an optimization hyperparameter that minimizes a first objective function is estimated. Next, a processing of updating a parameter using a second optimization method is executed on the basis of the acquired grouped uncoupled data and grouped size comparison data and the estimated optimization hyperparameter, and an optimization parameter that minimizes a second objective function is estimated. Finally, the estimated optimization parameter is outputted.