Encoding Model for Causal Feature Extraction in Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine training models struggle to accurately grasp causal relationships between measurement data and correct answer data, leading to decreased estimation accuracy when applied to different scenarios, such as determining hallux valgus in both males and females using sensor data.
Innovation Solution
A training device that includes an acquisition unit, a feature amount calculation unit, a model construction unit, and a training processing unit, which acquires raw data and correct answer data, calculates feature amounts, constructs encoding and estimation models, and trains them to match correct answer data based on the relationship between the code and feature amounts, ensuring accurate causal relationship understanding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If low-dimensional observation data processed up to a predetermined intermediate layer is transmitted from the instrument to a device, then the amount of data at the time of transmitting data can be reduced, but the analysis using the dimensionally reduced data may cause a decrease in estimation accuracy
Solution Approach 1:
The patent applies preliminary action by pre-processing raw data to extract feature amounts and generating codes that capture essential characteristics before transmission. The encoding model is trained in advance to convert raw data into compact code representations that preserve causal relationships, allowing the instrument to transmit reduced data while maintaining estimation accuracy at the analysis device
Solution Approach 2:
The patent introduces an intermediary encoding model that acts as a mediator between the instrument and the analysis device. This encoding model transforms raw data into feature-based codes that serve as an intermediate representation, preserving essential information while reducing data volume. The code contains extracted feature amounts that maintain the causal relationship needed for accurate estimation
2Adaptability or versatility
If a model trained by machine training is used to make determination, then the model can grasp a correlation between measurement data and correct answer data, but the model cannot grasp a causal relationship between measurement data and correct answer data
Solution Approach 1:
The patent applies parameter changes by transforming the model training approach from direct correlation learning to feature-based causal learning. Instead of training the model to directly correlate raw data with answers, the system extracts feature amounts as intermediate parameters and trains the encoding model to capture causal relationships through these features. This parameter transformation enables the model to understand causality rather than merely correlation
Data Source
AI summary
A training device includes an acquisition unit that acquires a data set of raw data and correct answer data, a feature amount calculation unit that calculates a feature amount using the raw data, a model construction unit that constructs an encoding model that outputs a code related to the feature amount in response to an input of the raw data and an estimation model that outputs an estimation result related to the raw data in response to an input of the code, and a training processing unit that trains the encoding model and the estimation model in such a way that the estimation result matches the correct answer data based on a relationship between the code and the feature amount.