湖北民族大学学报（自然科学版）2026，Vol.44Issue(2)：228-233,6.DOI:10.13501/j.cnki.42-1908/n.2026.06.007
基于AME-MoE模型的微生物浓度检验方法
Microbial Concentration Testing Method Based on the AME-MoE Model
摘要
Abstract
To address the issues of strong lag,high cost,and poor interpretability in traditional microbial concentration test methods,a new method of microbial concentration test based on an artificial microbial ecosystem and mixture-of-experts(AME-MoE)model was proposed.The model decomposed dynamically the complex microbial communities into specific bacterial species response modeled by multiple lightweight expert submodels,and introduced a Transformer to enhance the model's capture of nonlinear relationships,ultimately generating microbial concentration detection results.The research demonstrated that,referring to colony forming unit(CFU),the root mean square error of AME-MoE model was as low as 0.15lg(CFU/g),with a metabolic flux detection accuracy of 88.7%.Additionally,the single-sample inference latency time was only 22 ms,and the model exhibited excellent generalization capabilities in scenarios with data scarcity and in a noisy environment.AME-MoE model achieved high-precision,low-latency,and strongly generalizable microbial concentration intelligent detection,providing a decision-making tool with practical application value for hospital infection control,food safety,and real-time monitoring of environmental microbial risks.关键词
人工微生物生态系统/混合专家系统/微生物浓度检验/模型优化/智能决策Key words
artificial microbial ecosystem/mixture-of-expert system/microbial concentration inspection/model optimization/intelligent decision-making分类
信息技术与安全科学引用本文复制引用
张贤,徐建飞..基于AME-MoE模型的微生物浓度检验方法[J].湖北民族大学学报（自然科学版）,2026,44(2):228-233,6.基金项目
安徽省教学研究项目(2022jyxm619). （2022jyxm619）
