Pruning Filter in Filter
Authors: Fanxu Meng, Hao Cheng, Ke Li, Huixiang Luo, Xiaowei Guo, Guangming Lu, Xing Sun
NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through extensive experiments, we demonstrate that SWP is more effective compared to the previous FP-based methods and achieves the state-of-art pruning ratio on CIFAR-10 and Image Net datasets without obvious accuracy drop. Code is available at this url. |
| Researcher Affiliation | Collaboration | Fanxu Meng1,2 , Hao Cheng2 , Ke Li2, Huixiang Luo2, Xiaowei Guo2, Guangming Lu1 , Xing Sun2 1 Harbin Institute of Technology, Shenzhen, China 2 Tencent Youtu Lab, Shanghai, China |
| Pseudocode | No | The paper includes mathematical equations and diagrams illustrating the process (Figure 3, Figure 5), but it does not contain formal pseudocode or a clearly labeled algorithm block. |
| Open Source Code | Yes | Code is available at this url. |
| Open Datasets | Yes | Datasets and Models: CIFAR-10 [40] and Image Net [41] are two popular datastes and are adopted in our experiments. CIFAR-10 dataset contains 50K training images and 10K test images for 10 classes. Image Net contains 1.28 million training images and 50K test images for 1000 classes. |
| Dataset Splits | No | The paper specifies training and testing data. For CIFAR-10, it mentions '50K training images and 10K test images'. For ImageNet, it describes training with 'randomly crop a 224 224 area' and testing with 'center crop of 224 224 pixels'. However, it does not explicitly define a separate validation set split or its size/percentage for either dataset. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for experiments, such as CPU or GPU models, memory, or cloud computing instance types. |
| Software Dependencies | No | The paper mentions following 'the official Py Torch implementation' for Image Net experiments but does not specify the version number of PyTorch or any other software dependencies. |
| Experiment Setup | Yes | For CIFAR-10, the model was trained for 160 epochs with a batch size of 64. The initial learning rate is set to 0.1 and divide it by 10 at the epoch 80 and 120. For Image Net, we follow the official Py Torch implementation 1 that train the model for 90 epochs with a batch size of 256. The initial learning rate is set to 0.1 and divide it by 10 every 30 epochs. α is set to 1e-5 in (5) and the threshold δ is set to 0.05. |