论文信息 - Remote Sensing Image Target Recognition Based on Pruned Deep Neural Network Models

Remote Sensing Image Target Recognition Based on Pruned Deep Neural Network Models

In this work, pruned deep neural network models have been used for remote sensing image target recognition to reduce the storage resources. This model compression procedure including two parts: first, pre-train the model for remote sensing image target recognition. Second, compression models by network pruning and weight clustering. The pruning procedure was implemented by learning the important connections in the network, and the weight clustering procedure was implemented by k-means algorithm. Experiments were performed with the AlexNet and VGG-16 networks on the NWPU-RESISC-10 datasets. Results show that the proposed method reduced the storage of AlexNet by 11.4×, from 240MB to 20MB and reduced the VGG-16 by 11× from 552MB to 50MB, with no significant loss of accuracy.

[1] Jürgen Schmidhuber,et al. Framewise phoneme classification with bidirectional LSTM and other neural network architectures , 2005, Neural Networks.

[2] Lorien Y. Pratt,et al. Comparing Biases for Minimal Network Construction with Back-Propagation , 1988, NIPS.

[3] J. A. Hartigan,et al. A k-means clustering algorithm , 1979 .

[4] Misha Denil,et al. Predicting Parameters in Deep Learning , 2014 .

[5] Babak Hassibi,et al. Second Order Derivatives for Network Pruning: Optimal Brain Surgeon , 1992, NIPS.

[6] David S. Touretzky,et al. Advances in neural information processing systems 2 , 1989 .

[7] Geoffrey E. Hinton,et al. Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[8] Song Han,et al. Deep Compression: Compressing Deep Neural Network with Pruning, Trained Quantization and Huffman Coding , 2015, ICLR.

[9] Nikhil S. Ketkar. Stochastic Gradient Descent , 2017 .

[10] Geoffrey E. Hinton,et al. ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[11] Yoshua Bengio,et al. Gradient-based learning applied to document recognition , 1998, Proc. IEEE.

[12] Trevor Darrell,et al. Caffe: Convolutional Architecture for Fast Feature Embedding , 2014, ACM Multimedia.

[13] Maurice K. Wong,et al. Algorithm AS136: A k-means clustering algorithm. , 1979 .

[14] Song Han,et al. Learning both Weights and Connections for Efficient Neural Network , 2015, NIPS.