Anomaly Detection for Tabular Data via Contrastive Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection methods for tabular data face challenges in accurately identifying outliers without prior information on data structure, especially when features are not independent, and require tuning between different datasets.
Innovation Solution
A method using two neural networks to learn a multi-variate mapping for tabular data by maximizing mutual information between subsets of features, employing contrastive learning to differentiate between in-distribution and out-of-distribution records through vector representations, allowing for efficient inference and classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If existing anomaly detection methods are applied to tabular data without prior information on data structure, then the methods can be generally applicable, but the accuracy of identifying outliers deteriorates when features are not independent
Solution Approach 1:
The patent introduces an intermediary component (the neural network-based anomaly detection system with contrastive learning) that learns the complex dependencies between features automatically. This intermediary translates raw tabular data with unknown dependency structures into meaningful anomaly scores, resolving the contradiction between general applicability and accuracy by bridging the gap through learned representations rather than direct analysis
2Reliability
If existing anomaly detection methods are used, then they can detect outliers, but they require tuning between different datasets which reduces stability
Solution Approach 1:
The patent employs parameter changes through contrastive learning with temperature scaling. By introducing a temperature parameter that controls the sharpness of the contrastive loss function, the system adapts to different datasets without extensive manual tuning. This parameter adjustment mechanism allows the same base model to maintain both anomaly detection capability and stability across diverse tabular datasets
3Measurement precision
If contrastive learning is applied to maximize mutual information between feature subsets, then accuracy and stability improve, but computational complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the feature space into multiple subsets and processing them through separate neural network branches. This segmentation allows the contrastive learning mechanism to work on manageable feature groups rather than the entire feature set at once, reducing computational complexity while maintaining the ability to capture complex dependencies for accurate anomaly detection
Data Source
AI summary
The disclosure comprises a method to improve machine learning models by cleaning training data using anomaly detection, as well as anomaly detection per se. The method considers the task of finding out-of-class samples in tabular data, where little may be safely assumed about the structure of the data. The method captures the structure of the samples of the single training class, by learning mappings that maximize the mutual information between each sample and a part that is masked out. The mappings are learned by employing a contrastive loss that considers only one sample at a time. Once learned, the disclosure may score a test sample by measuring whether the learned mappings lead to a small contrastive loss using the masked parts of this sample. The experiments show accuracy advantage in comparison to the literature using the same set of hyperparameters as the state of the art results across benchmarks.