The following instructions will guide you through the setup and running processes of our code.

pip install torch==2.1.1 torchvision==0.16.1
pip install causal-conv1d==1.1.1
pip install mamba-ssm

• config.py: Contains settings for training configurations.
• MambaTab.py: Contains the method-related code of MambaTab.
• supervised_mambatab.py: Code for vanilla supervised learning settings.
• feature_incremental.py: Code related to incremental feature settings.
• train_val.py: Manages the training and validation processes over epochs.
• utility.py: Includes functionality for data reading and preprocessing.
• Currently, our code is applicable for binary classification scenarios as all the datasets referenced in the paper involve binary classification tasks.

Data can be downloaded using the links provided in the paper. Please ensure the following format rules are met:

• The dataset should be in .csv format.
• The header row should be the first row in the .csv file.
• The target column should be the last column in the .csv file.
• Rename the file to data_processed.csv and place it in the datasets/X folder, where X can be dress, cylinder, etc.

Use the config.py file to set the necessary parameters for running the code. Additional model-related configurations can be modified in the MambaTab.py file.

• Vanilla Supervised Learning: Execute supervised_mambatab.py. Ensure you modify the config.py according to your needs as per the comments provided in the code. For example for credit approval (CA) dataset, one example run is given below:
• 
• Feature Incremental Learning: Run feature_incremental.py.
  Feel free to adjust parameters and explore different configurations to achieve the best results.

If you find this repo useful in your research, please consider citing our paper as follows:

@inproceedings{mambatab,
  title={{MambaTab}: A Plug-and-Play Model for Learning Tabular Data},
  author={Ahamed, Md Atik and Cheng, Qiang},
  booktitle={2024 IEEE Conference on Multimedia Information Processing and Retrieval (MIPR)},
  year={2024},
  organization={IEEE}
}