Local AI Simulation Using Chunked Data for Secure Model Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge in condition-based maintenance of mechanical systems is the difficulty in effectively utilizing AI algorithms due to the need to share sensitive user data with AI vendors, leading to inappropriate algorithm selection, communication errors, and the risk of data exposure, while lacking timely and appropriate data analysis.
Innovation Solution
A simulation apparatus and method that allows users to verify AI effectiveness using their own data through supervised training, employing a concept of 'chunks' for setting input and output data, enabling in-place analysis and reducing data exposure risks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user data is shared with AI vendors for algorithm training, then AI algorithm effectiveness can be improved, but data security and confidentiality are compromised
Solution Approach 1:
The patent introduces a simulation apparatus as an intermediary that enables users to perform supervised training locally using their own data. This mediator allows AI algorithm development and verification without requiring direct data sharing with external vendors, thus maintaining data security while achieving algorithm effectiveness.
Solution Approach 2:
The simulation apparatus creates a local copy of the AI training environment, allowing users to replicate the supervised training process using their own data sets. This copying approach eliminates the need to share sensitive data with external parties while still enabling algorithm development and validation.
2Object-affected harmful factors
If supervised training is performed using user's own data, then data security is improved, but the complexity of setting and executing training increases
Solution Approach 1:
The simulation apparatus is designed as a universal platform that integrates multiple functions: data loading, chunk configuration, model training, and result verification. This multi-functional design consolidates complex operations into a single unified system, reducing the overall complexity for users despite the sophisticated processes involved.
Solution Approach 2:
The patent segments the training data into manageable chunks with configurable parameters (number of rows, columns, and chunks). This segmentation approach simplifies the execution complexity by breaking down large data sets into smaller, more manageable units that can be processed systematically.
3Object-affected harmful factors
If data is analyzed in-place without sharing, then data security is improved, but the ability to verify AI effectiveness may be reduced
Solution Approach 1:
The simulation apparatus enables users to perform self-service AI algorithm verification using their own data. Users can load their data, configure training parameters, execute supervised training, and verify AI effectiveness locally without needing to share data with external vendors, thus maintaining both data security and verification accuracy.
Solution Approach 2:
The system implements feedback mechanisms that allow users to verify AI model performance by comparing predicted outputs with actual results using their own data. This local feedback loop ensures accurate verification of AI effectiveness while maintaining data confidentiality, as users can assess model performance without external data sharing.
Data Source
AI summary
A model setting unit executes setting related to a first chunk of input data and a second chunk of training data on a basis of loaded training-purpose data, and setting related to a third chunk of test input data and a fourth chunk of expected data on a basis of loaded test-purpose data, as well as setting related to calculation accuracy of the machine learning model. A model computing unit executes computations of training with the use of the machine learning model on a basis of the first chunk and the second chunk, and computations of prediction with use of the machine learning model on a basis of results of the training and the third chunk, as well as computations of a comparison between results of the prediction and the fourth chunk. Each chunk is a batch block of data in sequential feeding of data to the machine learning model.