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

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.