Dynamic Network Surgery for Efficient DNNs
Authors: Yiwen Guo, Anbang Yao, Yurong Chen
NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | The effectiveness of our method is proved with experiments. |
| Researcher Affiliation | Industry | Yiwen Guo Intel Labs China yiwen.guo@intel.com Anbang Yao Intel Labs China anbang.yao@intel.com Yurong Chen Intel Labs China yurong.chen@intel.com |
| Pseudocode | Yes | Algorithm 1 Dynamic network surgery: the SGD method for solving optimization problem (1): |
| Open Source Code | Yes | Code and some models are available at https://github.com/yiwenguo/Dynamic-Network-Surgery. |
| Open Datasets | Yes | MNIST is a database of handwritten digits and it is widely used to experimentally evaluate machine learning methods. ... In the final experiment, we apply our method to Alex Net [13], which wins the ILSVRC 2012 classification competition. |
| Dataset Splits | No | For the XOR problem, it states 'half of them were used as training samples and the rest as test samples', but no validation split is mentioned. For ImageNet, it mentions testing on the validation set ('test our compressed model on the validation set') but does not specify the train/validation/test split percentages or counts for reproduction. |
| Hardware Specification | No | The paper mentions 'GPU implementation of Caffe package', but does not provide specific details about the GPU models, CPU, or any other hardware used for the experiments. |
| Software Dependencies | No | The paper mentions 'all the reference models are trained by the GPU implementation of Caffe package [12]', but it does not specify a version number for Caffe or any other software dependencies. |
| Experiment Setup | Yes | For Le Net-5, 'train this model for 10,000 iterations'. For Alex Net, 'after 450K iterations of training (i.e., roughly 90 epochs). ... We run 320K iterations for the convolutional layers and 380K iterations for the fully connected layers, which means 700K iterations in total (i.e., roughly 140 epochs).' |