Comparison and Analysis of the Open-Source Frameworks for Deep Learning

Dong-sheng GAO, Yan-rong ZHAO, Jing GAO, Hao WANG

Abstract


Deep Learning is the hottest trend now in AI and Machine Learning. The paper introduces four mainstream open-source frameworks for deep learning, including Caffe, TensorFlow, CNTK, Torchnet. And the open-source frameworks for deep learning are analyzed and compared from the aspects of network and model capability, interface, model deployment, performance, cross-platform and distributed. Finally, four open-source frameworks for deep learning are simulated on the dataset, and the advantages and disadvantages of each framework and its applicability are summarized.

Keywords


Deep learning, Open-source frameworks, Performance comparison


DOI
10.12783/dtcse/mcsse2016/10975

Refbacks

  • There are currently no refbacks.