Conditional GAN for ICT Abnormal Data Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for constructing causal models in ICT systems face challenges in accurately estimating abnormal portions, particularly when abnormalities occur that are not covered by predefined rules or when sufficient data during abnormalities is difficult to collect.
Innovation Solution
A learning data generation apparatus that utilizes a conditional hostile generative network (CGAN) to generate learning data for constructing models that estimate abnormal portions in ICT systems. This apparatus learns parameters for a generator and discriminator within the CGAN using observation data during abnormalities, enabling the generation of diverse abnormal data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If chaos engineering is used to collect abnormality data, then the quantity of abnormality data is improved, but the variety of abnormality types is limited to only those that can be conceived by humans
Solution Approach 1:
The patent uses a generative adversarial network (GAN) to copy and synthesize abnormality data patterns from existing real abnormality data. The GAN learns the distribution characteristics of real abnormality data and generates synthetic abnormality data that mimics real scenarios, thereby expanding both the quantity and variety of available training data without being limited by human-conceivable fault types
Solution Approach 2:
The patent applies parameter changes by using a conditional GAN that takes abnormality type labels as conditional parameters. By varying these conditional parameters, the system can generate diverse types of abnormality data while maintaining realistic patterns, thus achieving both quantity and variety improvements
2Reliability
If a causal model is constructed using expert knowledge and predefined rules, then the model can handle defined abnormalities, but it cannot correctly estimate abnormal portions when undefined abnormalities occur
Solution Approach 1:
The patent transitions from static expert-defined rules to a dynamic machine learning model that can adapt to new abnormality types. The GAN-based system learns patterns from data and can generalize to undefined abnormalities, providing both reliability for known cases and adaptability for unknown cases through its data-driven approach
Solution Approach 2:
The system uses unsupervised learning to automatically learn abnormality patterns from real abnormality data without requiring explicit expert labeling for each abnormality type. This self-learning capability enables the model to handle both defined and undefined abnormalities effectively
Data Source
AI summary
A learning data generation apparatus is a learning data generation apparatus generating learning data used to learn a model for estimating an abnormal portion of an ICT system. The learning data generation apparatus includes: a learning unit configured to learn parameters of a generator and a discriminator forming a conditional hostile generation network by using observation data during abnormality of the ICT system; and a generation unit configured to generate the learning data using the generator in which the learned parameters are set.