领域知识指导的模型重用  

Model reuse with domain knowledge

作　　者：吴西竹 周志华[1] 

机构地区：[1]南京大学计算机软件新技术国家重点实验室,南京210023

出　　处：《中国科学：信息科学》2017年第11期1483-1492,共10页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:61333014)资助项目

摘　　要：机器学习模型的生命周期通常比较短暂,大量的机器学习模型针对特定任务设计,在完成任务之后即失去使用价值.然而,一个精心设计和训练的模型通常更精炼地概括了训练数据中蕴含的知识.更进一步地,当无法获取原始训练数据时,已有的预训练模型就是仅剩的信息来源.本文提出了一种重用已有的预训练机器学习模型来辅助当前任务学习的框架,尤其适用于当前任务数据量不足的情形.该方法把已有模型视为黑盒,不需要已有模型的结构信息,并使用领域知识对已有模型进行筛选和结合.实验表明,在该框架下,通过对已有模型的重用,多个实际任务上的性能可以得到显著提升.The life spans of machine learning models are often short and a large number of models are wasted because they can only be applied to a specific task. However, a well-designed, carefully trained model contains learned knowledge from its task, which may be more concise than training data. Furthermore, when we have no access to training data, the trained model is the last remaining source of information. This study introduces a framework to reuse existing models trained in other tasks and help improve the model for the current task,especially when limited data is available for the current task. This framework incorporates high-level domain knowledge to combine existing models and treat them as black boxes, in order for them to be universal for complex models. Experiments on applying the framework to practical problems demonstrate that we can improve the performance on the current task by reusing existing models.

关 键 词：机器学习 模型重用 领域知识 环境变化 学件 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]