Refine or Represent: Residual Networks with Explicit Channel-wise Configuration

Authors: Yanyan Shen, Jinyang Gao

IJCAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results on the CIFAR10, CIFAR100 and Image Net datasets demonstrate that our proposed method can substantially improve the performance of conventional residual networks including Res Net, Res Ne Xt and SENet.
Researcher Affiliation Academia Yanyan Shen1 , Jinyang Gao2 1 Shanghai Jiao Tong University 2 National University of Singapore
Pseudocode No The paper does not contain any clearly labeled pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any specific links or explicit statements about the availability of open-source code for the described methodology.
Open Datasets Yes We evaluate the performance of Ro R using three image classification benchmarks: CIFAR-10, CIFAR-100 and Image Net. ... Dataset CIFAR10 ... #Train Images 50K ... Dataset CIFAR100 ... #Train Images 50K ... Dataset Image Net ... #Train Images 1280K
Dataset Splits No The paper states training and test image counts in Table 1, and Figure 2 shows 'validation' curves, implying a validation set was used. However, it does not explicitly provide the specific percentages or sample counts for the validation split, nor does it cite a predefined validation split.
Hardware Specification Yes All the models are implemented using Py Torch and trained with 2 GTX Titan XP GPUs.
Software Dependencies No The paper states 'All the models are implemented using Py Torch', but it does not provide a specific version number for PyTorch or any other software dependencies.
Experiment Setup Yes We train models on CIFAR dataset for 300 epochs and models on Image Net dataset for 80 epochs. For all the models, standard mini-batch SGD with momentum is used as the optimization method. The batch size is set to 128 and momentum is set to 0.9. All the parameters are initialized using the method described in [He et al., 2015]. The learning rate starts from 0.01 and is reduced by 10x when 50% and 75% of the epochs are finished.