Model Transferability with Responsive Decision Subjects

This repository accompanies the paper Model Transferability with Responsive Decision Subjects accepted by ICML 2023 -- Yatong Chen, Zeyu Tang, Kun Zhang, Yang Liu.

Abstract:

Given an algorithmic predictor that is accurate on some source population consisting of strategic human decision subjects, will it remain accurate if the population respond to it? In our setting, an agent or a user corresponds to a sample $(X,Y)$ drawn from a distribution $\mathcal{D}$ and will face a model $h$ and its classification result $h(X)$. Agents can modify $X$ to adapt to $h$, which will incur a distribution shift on $(X,Y)$.
Our formulation is motivated by applications where the deployed machine learning models are subjected to human agents, and will ultimately face responsive and interactive data distributions. We formalize the discussions of the transferability of a model by studying how the performance of the model trained on the available source distribution (data) would translate to the performance on its induced domain. We provide both upper bounds for the performance gap due to the induced domain shift, as well as lower bounds for the trade-offs that a classifier has to suffer on either the source training distribution or the induced target distribution. We provide further instantiated analysis for two popular domain adaptation settings, including covariate shift and target shift.

Guideline

Synthetic experiments using simulated data:

The result for synthetic experiments using simulated data is provided in the Jupyter notebook named 'Model-Transferability-ICML23.ipynb'. Detailed dscriptions of the data generating process can be found in Section 6 of the paper. Running the notebook will reproduce Figure 4.

Citation

If you want to cite our paper, please cite the following format:

@article{chen2023model,
  title={Model Transferability With Responsive Decision Subjects},
  author={Chen, Yatong and Tang, Zeyu and Zhang, Kun and Liu, Yang},
  booktitle={International Conference on Machine Learning},
  organization={PMLR}
  year={2023}
}