Unsupervised Dynamic Feature Selection for Robust Latent Spaces in Vision Tasks

2025-10-02

AI Summary

Visual representations are often corrupted by noise and task-irrelevant features in latent space, degrading model robustness and generalization. To address this, we propose an unsupervised dynamic feature selection method that operates during forward propagation: it performs instance-wise importance scoring and applies a lightweight gating mechanism to selectively retain discriminative features—entirely without supervision or ground-truth labels. This is the first work to introduce unsupervised dynamic feature selection for latent-space representation optimization, leading to substantially improved representation quality. Extensive experiments on image clustering and generation tasks demonstrate consistent performance gains across multiple benchmark datasets, along with enhanced generalization capability. Crucially, the method incurs only negligible computational overhead, preserving efficiency while significantly boosting representation fidelity.

Technology Category

• Machine Learning: Dimensionality Reduction/Feature Selection
• Computer Vision: Representation Learning for Vision
• Search and Optimization: Learning to Search

Application Category

• 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
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Latent representations are critical for the performance and robustness of machine learning models, as they encode the essential features of data in a compact and informative manner. However, in vision tasks, these representations are often affected by noisy or irrelevant features, which can degrade the model's performance and generalization capabilities. This paper presents a novel approach for enhancing latent representations using unsupervised Dynamic Feature Selection (DFS). For each instance, the proposed method identifies and removes misleading or redundant information in images, ensuring that only the most relevant features contribute to the latent space. By leveraging an unsupervised framework, our approach avoids reliance on labeled data, making it broadly applicable across various domains and datasets. Experiments conducted on image datasets demonstrate that models equipped with unsupervised DFS achieve significant improvements in generalization performance across various tasks, including clustering and image generation, while incurring a minimal increase in the computational cost.

Problem

• Removing noisy features from latent representations in vision tasks
• Enhancing model generalization without labeled data supervision
• Improving clustering and generation performance with minimal computational overhead

Innovation

• Dynamic Feature Selection removes redundant image features
• Unsupervised framework operates without labeled data
• Enhances latent spaces for clustering and generation