Pytorch implementation of LeNo-ResNet proposed in:
Bharat Mahaur et al. "Improved Residual Network Based on Norm-Preservation for Visual Recognition." Neural Networks 2022.

Please find the paper here: https://doi.org/10.1016/j.neunet.2022.10.023.

Install PyTorch and ImageNet dataset following the official PyTorch ImageNet training code.

A fast alternative (without installing PyTorch and other deep learning libraries) is to use NVIDIA-Docker.

To train a model (for instance, LeNo-ResNet with 50 layers) using DataParallel run main.py; you need to provide result_path (the directory path where to save the results and logs) and the --data (the path to the ImageNet dataset):

result_path=/your/path/to/save/results/and/logs/ mkdir -p ${result_path} python main.py \ --data /your/path/to/ImageNet/dataset/ \ --result_path ${result_path} \ --arch lenoresnet \ --model_depth 50

To train using Multi-processing Distributed DataParallel; follow the instructions in the official PyTorch ImageNet training code.

The gradient norm ratios for ResNet, pre-act ResNet, and LeNo-ResNet over 200-layers depth network:

If you use this code, please cite our paper:

@article{mahaur2022improved, title={Improved Residual Network Based on Norm-Preservation for Visual Recognition}, author={Mahaur, Bharat and others}, journal={Neural Networks}, year={2022}, publisher={Elsevier} } 

Please contact bharatmahaur@gmail.com for any further queries.

This code is released under the Apache 2.0 License.