Multi-Adapter PPO: A Cross-Attention Enhanced Wavelength Selection Framework for LIBS Quantitative Analysis

2026-06-15

AI Summary

This study addresses the challenges of wavelength selection and the trade-off between accuracy and efficiency in quantitative analysis using laser-induced breakdown spectroscopy (LIBS), which arise from high-dimensional spectral data. To this end, the authors propose a novel reinforcement learning–driven feature selection framework that, for the first time, integrates multi-adapter proximal policy optimization (PPO) with a cross-attention mechanism. This approach leverages the multi-adapter architecture and cross-attention to effectively model complex inter-spectral relationships, enabling interpretable yet highly efficient and precise wavelength selection. Experimental results on steel and coal datasets demonstrate that the proposed method achieves state-of-the-art performance in LIBS-based quantitative analysis, yielding an average improvement of 28.4% in composite scores and a 45.2% increase in prediction accuracy.

Abstract

Laser-induced breakdown spectroscopy (LIBS) quantitative analysis faces critical challenges in wavelength selection due to high-dimensional spectral data and the fundamental trade-off between prediction accuracy and feature efficiency. This paper presents a novel Multi-Adapter PPO framework that transforms wavelength selection into a reinforcement learning problem, leveraging cross-attention mechanisms and multiple specialized adapters to capture complex spectral relationships. Our approach outperforms traditional Particle Swarm Optimization (PSO) by an average of 28.4\% in comprehensive score and 45.2\% in prediction accuracy across steel and coal datasets. The proposed method demonstrates superior performance in balancing prediction accuracy with feature efficiency, achieving state-of-the-art results in LIBS quantitative analysis while maintaining interpretability and computational efficiency. We released our code and dataset here: https://github.com/Hflying/MAPPO

Problem

• wavelength selection
• LIBS quantitative analysis
• feature efficiency
• prediction accuracy
• high-dimensional spectral data

Innovation

• Multi-Adapter PPO
• cross-attention
• wavelength selection
• reinforcement learning
• LIBS quantitative analysis