Training Data Constraint Filtering Using Metadata Compliance Rules
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The collection and use of application data for model training often occur non-specifically, neglecting data compliance requirements, leading to increased violations of regulatory standards and potential penalties.
Innovation Solution
A system for constraint-based training data generation that includes metadata annotation and compliance enforcement, utilizing software-defined networking and data compliance as code (DCaC) to ensure data handling adheres to regulatory standards across multi-cloud and edge infrastructures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If application data is collected and used for model training without specific compliance checks, then model training productivity is improved, but data compliance reliability deteriorates
Solution Approach 1:
The system performs preliminary compliance assessment before data is used for model training. Data compliance constraints are evaluated and enforced in advance through metadata annotation and constraint-based filtering, ensuring that only compliant data enters the training pipeline, thus maintaining reliability without sacrificing productivity
Solution Approach 2:
The patent introduces an intermediary compliance assessment layer between data collection and model training. This intermediary system uses metadata annotations and constraint evaluation to filter and validate data, acting as a mediator that ensures compliance requirements are met while enabling efficient model training with validated data
2Reliability
If comprehensive data compliance constraints are enforced, then data compliance reliability is improved, but system complexity increases
Solution Approach 1:
The system changes the parameter representation of compliance constraints by using metadata annotations attached to data elements. Instead of complex rule-based systems, compliance requirements are expressed as configurable parameters and constraints that can be automatically evaluated, reducing system complexity while maintaining comprehensive compliance enforcement
Solution Approach 2:
The patent segments compliance enforcement into modular components: metadata annotation, constraint definition, constraint evaluation, and filtering. Each component handles a specific aspect of compliance, making the overall system more manageable and less complex while achieving comprehensive compliance coverage
3Reliability
If metadata annotation and constraint evaluation are implemented for all application data, then data compliance reliability is improved, but processing time increases
Solution Approach 1:
The system applies partial action by evaluating compliance constraints only for data that requires it, based on metadata annotations. Not all application data undergoes full constraint evaluation - only data with relevant compliance requirements is processed, reducing overall processing time while maintaining compliance reliability for affected data
Data Source
AI summary
In one embodiment, a device may receive a request for training data that is based on application data generated by an application executed at a data collection node, wherein the application data is associated with metadata identifiers. The device may determine one or more training data constraints that restrict use of the application data as training data. The device may generate the training data in part by excluding application data of a particular type from being included in the training data based on a match between its metadata identifier and the one or more training data constraints. The device may provide the training data to be used to train a machine learning model.