Deep Learning for Two-Sided Matching

This folder containts the implementation of the paper: Deep Learning for Two-Sided Matching

Getting Started

The code is written in python3 and requires the following packages

• Numpy
• Numba
• Matplotlib
• PyTorch

Running the Experiments

We implement the following architectures:

Architecture	Train filename
MLP	train_MLP.py
CNN	train_CNN.py

To run MLP, do

python <train_filename> -n <num_agents> -p <truncation_probability> -c <correlation_probability> -l <lambda>

To change other hyperparameters, visit the corresponding file and modify the Args class.
The logfiles and the saved models can be found in experiments/ folder

Citing the Project

Please cite our work if you find our code/paper is useful to your work.

@article{ravindranath2021deep,
  title={Deep learning for two-sided matching},
  author={Ravindranath, Sai Srivatsa and Feng, Zhe and Li, Shira and Ma, Jonathan and Kominers, Scott D and Parkes, David C},
  journal={arXiv preprint arXiv:2107.03427},
  year={2021}
}