Snapshot-Ensemble-Colorectal-Cancer

Implementation of our paper SnapEnsemFS: A Snapshot Ensembling-based Deep Feature Selection Model for Colorectal Cancer Histological Analysis published in Scientific Reports, Nature (2023).

Overall Workflow

Requirements

To install the required dependencies run the following in command prompt: pip install -r requirements.txt

Running the codes:

Required directory structure:

+-- data
|   +-- .
|   +-- train
|   +-- val
+-- PSO.py
+-- __init__.py
+-- main.py
+-- model.py

Then, run the code using the command prompt as follows:

python main.py --data_directory "data"

Available arguments:

• --epochs: Number of epochs of training. Default = 100
• --learning_rate: Learning Rate. Default = 0.0002
• --batch_size: Batch Size. Default = 4
• --momentum: Momentum. Default = 0.9
• --num_cycles: Number of cycles. Default = 5

Citation

If you find our paper useful for your research, consider citing us:

@article{chattopadhyay2023snapensemfs,
  title={SnapEnsemFS: a snapshot ensembling-based deep feature selection model for colorectal cancer histological analysis},
  author={Chattopadhyay, Soumitri and Singh, Pawan Kumar and Ijaz, Muhammad Fazal and Kim, SeongKi and Sarkar, Ram},
  journal={Scientific Reports},
  volume={13},
  number={1},
  pages={9937},
  year={2023},
  publisher={Nature Publishing Group UK London}
}