Convolution with even-sized kernels and symmetric padding
Authors: Shuang Wu, Guanrui Wang, Pei Tang, Feng Chen, Luping Shi
NeurIPS 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, the efficacy of symmetric padding is validated in CIFAR10/100 [20] and Image Net [33] classification tasks, as well as CIFAR10, LSUN bedroom [44], and Celeb A-HQ [18] generation tasks. |
| Researcher Affiliation | Academia | Shuang Wu1, Guanrui Wang1, Pei Tang1, Feng Chen2, Luping Shi1 1Department of Precision Instrument, 2Department of Automation Center for Brain Inspired Computing Research Beijing Innovation Center for Future Chip Tsinghua University {lpshi,chenfeng}@mail.tsinghua.edu.cn |
| Pseudocode | No | The paper describes the steps for symmetric padding but does not provide structured pseudocode or an algorithm block. |
| Open Source Code | Yes | Our example code and models are available at https://github.com/boluoweifenda/CNN. |
| Open Datasets | Yes | In this section, the efficacy of symmetric padding is validated in CIFAR10/100 [20] and Image Net [33] classification tasks, as well as CIFAR10, LSUN bedroom [44], and Celeb A-HQ [18] generation tasks. |
| Dataset Splits | No | The paper mentions using well-known datasets like CIFAR10/100 and ImageNet, which typically have predefined splits. However, it does not explicitly state the specific training, validation, and test split percentages or sample counts used for reproduction. |
| Hardware Specification | Yes | In our reproduction, the training speeds on Titan XP for NASNet-A and Wide-Des Net are about 200 and 400 SPS, respectively. |
| Software Dependencies | No | The paper mentions deep learning frameworks like TensorFlow, Caffe, and PyTorch, and states that the implementation is in high-level Python and TensorFlow, but it does not specify any version numbers for these software components or other libraries. |
| Experiment Setup | Yes | The default settings for CIFAR classifications are as follows: We train models for 300 epochs with mini-batch size 64 except for the results in Table 2, which run 600 epochs as in [48]. We use a cosine learning rate decay [24] starting from 0.1 except for Dense Net tests, where the piece-wise constant decay performs better. The weight decay factor is 1e-4 except for parameters in depthwise convolutions. The standard augmentation [22] is applied and the α equals 1 in mixup augmentation. For Image Net classifications, all the models are trained for 100 epochs with mini-batch size 256. The learning rate is set to 0.1 initially and annealed according to the cosine decay schedule. We follow the data augmentation in [36]. Weight decay is 1e-4 in Res Net-50 and Dense Net-121 models, and decreases to 4e-5 in the other compact models. |