Temporal Segmentation for Consistent Machine Learning Explanations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for computing explanations of predictive machine learning outputs are inconsistent over time due to their reliance on temporal data, leading to uncertain and non-consistent explanations.
Innovation Solution
A method that involves temporally segmenting inputs to generate time windows, training models for each window, generating predictions, determining mode explanations, and reconciling explanations using a consistency factor to produce consistent explanations across different time windows.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional techniques are used to compute explanations of predictive outputs, then the explanations can be generated quickly, but the explanations are inconsistent and uncertain over time due to temporal dependencies in data
Solution Approach 1:
The patent segments temporal data into distinct time windows and trains separate models for each time window. This segmentation allows the system to capture temporal dependencies while maintaining consistency within each segment, resolving the contradiction between reliability and complexity by organizing the computational system into manageable temporal segments rather than attempting to process all temporal data simultaneously in a single model
2Adaptability or versatility
If data points that evolve over time are used in predictions, then the predictions can capture temporal patterns, but the explanations become highly uncertain and continuously changing
Solution Approach 1:
The patent implements a dynamic model selection mechanism where different models are selected and trained for different time windows based on the temporal characteristics of the data. This allows the system to adapt to changing temporal patterns while maintaining stable explanations within each time window, as each model is optimized for its specific temporal context rather than attempting to handle all temporal variations in a single static model
3Reliability
If a single model is used for predictions across all time periods, then the system is simpler to implement, but it cannot provide consistent explanations over time due to temporal data evolution
Solution Approach 1:
The patent divides the temporal data into multiple time windows and trains separate models for each segment. This segmentation enables the system to provide consistent explanations within each time window by capturing temporal dependencies specific to that period, while the modular structure allows for efficient parallel training and inference across segments, maintaining computational productivity despite using multiple models
Data Source
AI summary
A method for providing explanations of predictive outputs is disclosed. The method includes receiving, via an application programming interface, an input; temporally segmenting the input to generate a finite set of time windows; training machine learning models for each of the time windows; generating, by using each of the trained machine learning models, predictions for a target time based on the input; generating a set of common background data for each of the time windows based on the input; determining respective mode explanations for each of the time windows based on the corresponding set of common background data, the corresponding trained machine learning models, and the corresponding predictions; and determining reconciled explanations for a target prediction that corresponds to the target time based on the input and the respective mode explanations.