Sculpting Features from Noise: Reward-Guided Hierarchical Diffusion for Task-Optimal Feature Transformation

2025-05-21

AI Summary

Existing feature transformation methods face scalability bottlenecks in large combinatorial spaces: discrete search strategies suffer from poor scalability, while continuous optimization is prone to local optima. To address this, we propose RewardDiff—a reward-guided hierarchical diffusion generative framework. First, a variational autoencoder (VAE) constructs a compact latent space. Second, a latent-space diffusion model performs global and robust feature representation search under guidance from a performance evaluator (i.e., reward signal). Third, a semi-autoregressive decoder jointly models intra-feature dependencies and enables inter-feature parallel generation. RewardDiff is the first method to reformulate feature transformation as a reward-driven diffusion generative process. Evaluated on 14 benchmark datasets, it significantly outperforms state-of-the-art approaches in prediction accuracy and robustness, while achieving substantial improvements in both training and inference efficiency.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Feature Construction/Reformulation
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Feature Transformation (FT) crafts new features from original ones via mathematical operations to enhance dataset expressiveness for downstream models. However, existing FT methods exhibit critical limitations: discrete search struggles with enormous combinatorial spaces, impeding practical use; and continuous search, being highly sensitive to initialization and step sizes, often becomes trapped in local optima, restricting global exploration. To overcome these limitations, DIFFT redefines FT as a reward-guided generative task. It first learns a compact and expressive latent space for feature sets using a Variational Auto-Encoder (VAE). A Latent Diffusion Model (LDM) then navigates this space to generate high-quality feature embeddings, its trajectory guided by a performance evaluator towards task-specific optima. This synthesis of global distribution learning (from LDM) and targeted optimization (reward guidance) produces potent embeddings, which a novel semi-autoregressive decoder efficiently converts into structured, discrete features, preserving intra-feature dependencies while allowing parallel inter-feature generation. Extensive experiments on 14 benchmark datasets show DIFFT consistently outperforms state-of-the-art baselines in predictive accuracy and robustness, with significantly lower training and inference times.

Problem

• Overcoming limitations in Feature Transformation methods
• Navigating combinatorial spaces and avoiding local optima
• Generating task-optimal features efficiently and robustly

Innovation

• Uses VAE to learn compact feature latent space
• Employs LDM with reward guidance for optimization
• Decodes embeddings via semi-autoregressive structured generation