Knoop: practical enhancement of knockoff with over-parameterization for variable selection

2025-01-01

Machine-mediated learning

AI Summary

To address the challenge of model-agnostic variable selection in high-dimensional settings, this paper proposes Knoop—a novel knockoff-based method that introduces overparameterization into the knockoff framework for the first time, mitigating its construction difficulty and power degradation under small-sample regimes. Methodologically, Knoop leverages deep generative models to learn adaptable knockoff features, incorporates an adaptive thresholding strategy, provides theoretical finite-sample false discovery rate (FDR) control, and enhances randomization stability. Extensive experiments across multivariate linear regression, generalized linear models, and nonlinear scenarios demonstrate that, while strictly controlling FDR at level α = 0.1, Knoop achieves average power gains of 23%–41% over existing knockoff methods, with manageable computational overhead. These advances significantly improve the practicality and reliability of knockoffs in real-world high-dimensional inference tasks.

Technology Category

• Machine Learning: Kernel Methods
• Reasoning under Uncertainty: Stochastic Optimization
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for incomplete, noisy, or partially observed Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Problem

• Feature Selection
• Model Accuracy
• Efficiency

Innovation

• Variable Selection
• Anomaly Detection
• Regression Model