Data Distributional Properties As Inductive Bias for Systematic Generalization

This repository contains the code to replicate the experiments and results presented in the paper:

"Data Distributional Properties As Inductive Bias for Systematic Generalization" Felipe del Rio, Alain Raymond-Saez, Daniel Florea, Rodrigo Toro Icarte, Julio Hurtado, Cristian B. Calderon, Alvaro Soto Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), June 2025.

Overview

The code in this repository implements the methods and experiments described in the paper, focusing on leveraging data distributional properties as inductive biases to improve systematic generalization in machine learning models.

Features

• Implementation of the proposed inductive bias methods.
• Scripts for training and evaluation.
• Reproducible experiments as described in the paper.

Requirements

• Python 3.8+
• Required libraries are listed in requirements.txt.

Usage

1. Clone the repository:

   git clone https://github.com/fdelrio89/data-systematic.git
   cd data-systematic

2. Install dependencies:

   pip install -r requirements.txt

3. Run experiments:

   python run_training.py

Citation

If you use this code in your research, please cite the paper:

@InProceedings{del_Rio_2025_CVPR,
     author    = {del Rio, Felipe and Raymond-Saez, Alain and Florea, Daniel and Icarte, Rodrigo Toro and Hurtado, Julio and Calderon, Cristian B. and Soto, Alvaro},
     title     = {Data Distributional Properties As Inductive Bias for Systematic Generalization},
     booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)},
     month     = {June},
     year      = {2025},
     pages     = {25590-25601}
}

Contact

For questions or issues, please contact fidelrio at uc dot cl.