Programming frameworks for deep learning algorithms

Abstract

In recent years,deep learning algorithms became increasingly pervasive.It has drew extensive attentions from both researchers and industries,as it achieves very promising results on many applications of various fields.More and more researchers began to use deep learning algorithms to solve practical problems (e.g.,image classification,image recognition,speech recognition,and natural language processing).Many deep learning frameworks and libraries were proposed so that researchers can develop new deep learning algorithms in a more convenient fashion.These frameworks and libraries were different in many aspects (e.g.,design principles and abstraction).Firstly,several pervasive deep learning frameworks were introduced,and then the critical issue of designing such frameworks was analyzed.At last,the future challenges of designing deep learning frameworks were discussed.The study provides ideas and directions for future design.

Keywords

• machine learning;programming framework;big data