Bias Parameter Learning for Classification Prioritization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning algorithms, including inverse reinforcement learning, face challenges in intentionally biasing classification results, as the objective function does not inherently account for the degree of bias in classification outcomes, leading to equal treatment of normal and abnormal data, making it difficult to prioritize one classification over the other.
Innovation Solution
A learning device and method that incorporates a bias parameter into the objective function to adjust the scores of classification results, using inverse reinforcement learning to optimize logistic regression weights and estimate the bias parameter, allowing for intentional biasing of classification outcomes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional objective functions are used in machine learning, then classification results are treated equally, but the ability to intentionally bias classification results is lost
Solution Approach 1:
The patent introduces a bias parameter β into the objective function that can adjust the weight of different classification outcomes. By changing this parameter, the system can intentionally bias classification results toward specific outcomes (e.g., prioritizing detection of abnormal cases over normal cases), thereby resolving the contradiction between maintaining equal treatment and enabling intentional biasing.
Solution Approach 2:
The patent segments the objective function into multiple terms, each corresponding to different classification outcomes (true positive, false positive, true negative, false negative). Each term is weighted by a bias parameter, allowing independent control over the importance of each outcome type. This segmentation enables flexible biasing without requiring complete redesign of the objective function.
2Ease of operation
If bias parameters are introduced to prioritize certain classification results, then classification bias control is improved, but the complexity of learning increases
Solution Approach 1:
The patent employs inverse reinforcement learning where the system automatically learns the optimal bias parameters from expert decision-making data without requiring manual specification. The learning algorithm infers the implicit bias preferences from observed expert behavior, thereby providing ease of operation while managing learning complexity through automation.
Solution Approach 2:
The patent uses expert decision-making history data as feedback to iteratively refine the bias parameters. The system compares its classifications against expert decisions and adjusts the bias parameters accordingly, creating a closed-loop learning process that improves control over classification bias while managing complexity through data-driven adaptation.
Data Source
AI summary
An input means 81 accepts input of an extended objective function, in which each term indicative of a score of each classification result in an objective function of classification analysis is multiplied by a bias parameter as a parameter indicative of a degree of bias of the score of each classification result concerned. An optimization means 82 optimizes a logistic regression weight in the extended objective function. An estimation means 83 estimates the bias parameter by inverse reinforcement learning using the extended objective function of logistic regression to which the optimized weight is set.