Towards Convolutional Neural Networks Compression via Global Error Reconstruction

Authors: Shaohui Lin, Rongrong Ji, Xiaowei Guo, Xuelong Li

IJCAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed GER method is evaluated on the ILSVRC2012 image classification benchmark, with implementations on two widely-adopted convolutional neural networks, i.e., the Alex Net and VGGNet-19. Comparing to several state-of-the-art and alternative methods of CNN compression, the proposed scheme has demonstrated the best rate-distortion performance on both networks.
Researcher Affiliation Collaboration Shaohui Lin1,2, Rongrong Ji1,2 , Xiaowei Guo3, Xuelong Li4 1Fujian Key Laboratory of Sensing and Computing for Smart City, Xiamen University, 361005, China 2School of Information Science and Engineering, Xiamen University, 361005, China 3Best Image, Tencent Technology (Shanghai) Co.,Ltd, China 4Xi an Institute of Optics and Precision Mechanics, Chinese Academy of Sciences, Xi an, China
Pseudocode Yes Algorithm 1 Alternating optimization layer-by-layer; Algorithm 2 Global error Reconstruction for compressing CNN
Open Source Code No The paper mentions implementing GER on AlexNet and VGGNet-19, and training using Caffe, but does not provide concrete access to the source code for their proposed GER method.
Open Datasets Yes We test the proposed GER based CNN compression on the ILSVRC2012 image classification benchmark. The dataset contains more than 1 million training images from 1,000 object classes.
Dataset Splits Yes The dataset contains more than 1 million training images from 1,000 object classes. It also has a validation set of 50,000 images, where each object class contains 50 images. We randomly select 100,000 images (100 from each class) from the training set for training, and test on the validation set.
Hardware Specification Yes The compressed networks are trained using Caffe [Y. Jia and Darrell, 2014] and run on NVIDIA GTX TITAN X graphics card with 12GB.
Software Dependencies No The paper mentions using 'Caffe' for training, but does not provide a specific version number or other software dependencies with versions.
Experiment Setup Yes The learning rate starts at 0.01 and is halved every 10-epochs; the weight decay is set to 0.0005 and the momentum is set to 0.9.