Human Uncertainty Inference Using Proxy Ensemble Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in quantifying and inferring human uncertainty due to low accessibility and interpretability, particularly in machine learning models, which hinders efficient learning and decision-making processes.
Innovation Solution
A computer system utilizing a proxy ensemble network (PEN) to estimate and infer predictive uncertainty for individual humans, enabling accurate uncertainty range estimation without repeated measurements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a machine learning model uses sampling to calculate uncertainty, then accessibility is improved, but measurement precision deteriorates for human uncertainty
Solution Approach 1:
The patent creates a computational model (proxy ensemble network) that copies and simulates human uncertainty patterns. By training the model on human uncertainty data, it reproduces human-like uncertainty estimates, enabling machine systems to access human uncertainty measurements efficiently without requiring repeated human sampling, thus improving accessibility while maintaining measurement precision.
Solution Approach 2:
The patent performs preliminary training of the proxy ensemble network using human uncertainty data before actual deployment. This preliminary action captures human uncertainty patterns in advance, allowing the model to quickly infer uncertainty for new data points without requiring real-time human input or repeated sampling, thereby improving both accessibility and measurement precision.
2Measurement precision
If repeated measurements are performed to accurately infer human uncertainty, then measurement precision is improved, but productivity deteriorates
Solution Approach 1:
The proxy ensemble network is trained in advance on human uncertainty data, capturing uncertainty patterns beforehand. This preliminary action allows the model to rapidly infer uncertainty for new data points using the learned patterns, eliminating the need for repeated measurements while maintaining high measurement precision and significantly improving productivity.
Solution Approach 2:
The patent replaces the mechanical process of repeated human measurements with a computational inference process. The trained proxy ensemble network substitutes for repeated sampling, using learned patterns to quickly estimate uncertainty without requiring actual repeated measurements, thereby improving productivity while maintaining measurement precision.
3Ease of operation
If a proxy ensemble network is trained on human uncertainty data, then accessibility is improved, but device complexity increases
Solution Approach 1:
The patent extracts the essential uncertainty patterns from human data and isolates them into a dedicated proxy ensemble network. By separating the uncertainty inference function from the main system and encapsulating it in a trained model, the system achieves improved accessibility while managing complexity through modular design, where the complex training process is performed once offline.
Data Source
AI summary
Provided is a computer system and method for inferring a human uncertainty that may estimate a predictive uncertainty of a human about input data based on a proxy ensemble network configured for each individual human and may infer an uncertainty range including the predictive uncertainty for the human. The proxy ensemble network may be configured using uncertainty measurement values for the respective data items evaluated by the human.