Automation and Feature Selection Enhancement with Reinforcement Learning (RL)
Sumana Sanyasipura Nagaraju
TL;DR
This work surveys reinforcement learning–driven approaches for feature selection, representation, and transformation, emphasizing multi-agent and interactive paradigms that navigate high-dimensional spaces efficiently. It introduces and contrasts a range of methodologies—from MARLFS and IRFS to Monte Carlo-based, dual-agent, and hierarchical frameworks—along with bandit-based strategies and graph-embedding techniques for state representation. The key contributions include diverse RL architectures, state representations via convolutional auto-encoders and graphs, and practical strategies like early stopping and external trainer guidance to balance accuracy and computational cost. The findings suggest RL-based feature selection can outperform traditional methods in scalability and interpretability, with broad applicability to domains such as medicine, finance, and robotics. This points to a path toward more automated, robust, and efficient feature engineering in real-world ML systems.
Abstract
Effective feature selection, representation and transformation are principal steps in machine learning to improve prediction accuracy, model generalization and computational efficiency. Reinforcement learning provides a new perspective towards balanced exploration of optimal feature subset using multi-agent and single-agent models. Interactive reinforcement learning integrated with decision tree improves feature knowledge, state representation and selection efficiency, while diversified teaching strategies improve both selection quality and efficiency. The state representation can further be enhanced by scanning features sequentially along with the usage of convolutional auto-encoder. Monte Carlo-based reinforced feature selection(MCRFS), a single-agent feature selection method reduces computational burden by incorporating early-stopping and reward-level interactive strategies. A dual-agent RL framework is also introduced that collectively selects features and instances, capturing the interactions between them. This enables the agents to navigate through complex data spaces. To outperform the traditional feature engineering, cascading reinforced agents are used to iteratively improve the feature space, which is a self-optimizing framework. The blend of reinforcement learning, multi-agent systems, and bandit-based approaches offers exciting paths for studying scalable and interpretable machine learning solutions to handle high-dimensional data and challenging predictive tasks.