多示例学习是与监督学习、非监督学习和强化学习并列的第四类学习框架，目前已广泛应用于药物设计、图像搜索等领域,并已获得很好的效果。在多示例学习中，训练样本是由多个示例组成的包，包是有概念标记的，但示例本身却没有概念标记，学习的目的是预测新包的类别。-Multi-instance learning and supervised learning, unsupervised learning and reinforcement learning tied for the fourth-class learning fr a mework, has been widely used in the field of drug design, image search, and have obtained very good results. Multi-instance learning, the training sample is composed by a number of example package, the package is the concept of mark, but the sample itself has no concept of mark, the purpose of study is to predict the category of the new package.