一种适应大数据处理要求的深层学习模型

提出基于参数更新与结构更新的深层学习模型.将向量空间表示扩展到张量空间表示,解决数据结构表示的局限性,能很好地表示非结构化、半结构化等高维数据.改进后的模型具备保留原有训练参数、数据等信息能力,不需要再次训练原数据集,并且能根据新进数据进行参数调整或结构调整更新,从而实现对新进数据的学习.实验结果表明,该模型具有很好的...

A DEEP LEARNING MODEL FOR BIG DATA PROCESSING

This paper proposes a deep learning model based on parameter updating and structure updating.It extended vector space representation to tensor space representation and solved the limitation of data structure representation,and could well express unstructured,semi-structured and other high-dimensional data.The improved model can retain the original training parameters,data and other information without the need to train the original data set again,and can adjust the parameters or update the structure according to the new data,so as to realize the learning of the new data.The experimental results show that the model has good adaptability and maintenance,and can meet the requirements of fast change and complex type of data under the background of big data,and has good practicability of expansion.