Encoder Feature Segmentation for Explainable Multi-Task Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Machine learning models used in automated driving and other applications have low explainability, making it difficult to analyze computation content and leading to potential malfunctions and decreased accuracy due to the complexity and randomness of computation parameters.
Innovation Solution
A two-step training process for a learning model involving an encoder and multiple estimators is employed to ensure that different portions of feature amounts are highly relevant to specific estimation tasks, reducing interference and increasing explainability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If machine learning models are trained to execute estimation tasks, then the ability to perform regression or classification is improved, but the explainability of computation content deteriorates
Solution Approach 1:
The patent segments the computation parameters into multiple groups (first computation parameters and second computation parameters) corresponding to different estimation tasks. This segmentation allows the model to maintain high adaptability for executing multiple estimation tasks while improving explainability by associating specific parameter groups with specific tasks, making it easier to analyze and interpret computation content for each task independently.
2Adaptability or versatility
If the number of computation parameters is increased to handle more complicated estimation tasks, then the estimation capability is improved, but the difficulty of reading and analyzing computation content increases
Solution Approach 1:
The patent divides the large set of computation parameters into multiple segmented groups, where each group corresponds to a specific estimation task. This segmentation reduces the difficulty of analyzing computation content by organizing parameters in a structured manner, allowing users to focus on relevant parameter groups for each task without being overwhelmed by the entire complex parameter set.
Solution Approach 2:
The patent applies local quality by assigning specific characteristics to different parameter groups. Each group of computation parameters is associated with a specific estimation task and has localized properties that make it easier to interpret and analyze, while the overall model maintains high capability for complex estimation tasks through the combination of these specialized groups.
3Measurement precision
If computation parameters are adjusted to achieve desired estimation abilities, then the estimation accuracy is improved, but the consistency of computation content across different local solutions deteriorates
Solution Approach 1:
The patent segments computation parameters into task-specific groups, which stabilizes the composition of computation content for each task. By organizing parameters this way, the model achieves high estimation accuracy through optimized parameter adjustment while maintaining consistency in computation content across different local solutions, as each segment corresponds to a well-defined task with stable computational characteristics.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A model generation apparatus according to one aspect of the present invention executes: a first training step of training an encoder, a first estimator, and a second estimator so that, when training data of each of learning data sets is given to the encoder, a result of the first estimator executing a first estimation task matches first correct answer data, and a result of the second estimator executing a second estimation task matches second correct answer data; and a second training step of training the encoder so that, for each of the learning data sets, a correlation between a first portion of feature amounts and the second correct answer data is lowered, and a correlation between a second portion of the feature amounts and the first correct answer data is lowered.