深度学习的分区域工作机制、迁移学习的知识重用机制以及具有时序信息处理特性的Spiking脉冲发放机制是生物神经系统的三种重要的信息处理方式。目前对于深度时序模型的Transfer Learning方法的研究，大多借鉴生物神经系统的分区工作机制和知识重用机制，缺乏对神经元的时序信息表示方法的模拟，导致其很难有效的迁移复杂高维时序数据中蕴含的知识。本课题在深度时序Transfer Learning最新研究成果的基础上，利用具有时序表达能力的Spiking脉冲发放机制，探索既具有高维时空特征表示和知识迁移能力，又具有较强普适性的Spiking-Transfer Learning新模型和训练方法。这些理论与方法的研究可以大幅度降低深度时序模型训练所需的数据量且提升模型的时序处理能力，为机器学习在高维时序数据处理中的进一步应用奠定理论基础。

Transfer learning on deep models has achieved a lot of success because it adopts the deep learning strategy and the knowledge reuse mechanism from the biological neural systems. However, the existing deep transfer learning models ignore the spiking emitting mechanism, which is an important information representation approach in the biological neural systems. This leads to the failure of these models on transferring useful knowledge in the high-dimensional sequential data. To solve this problem, in this study, we explore a new transfer learning mechanism based on the third generation of neural networks, the Spiking neural networks. By which the transfer learning method can imitate the sequential data representation mechanism in the biological system, which could improve the sequential knowledge transfer capability of the models. The study on the spiking-transfer learning method can be applied to the high-dimensional sequential applications, and help to understand the cognitive science and computational neuroscience in the machine learning in a high level.