# Adversarial Weighting for Domain Adaptation in Regression

**Authors:** Antoine de Mathelin, Guillaume Richard, Mathilde Mougeot, Nicolas Vayatis  
**arXiv:** [2006.08251](https://arxiv.org/abs/2006.08251)  
**Date:** 2020-06-15  
**Summary page:** https://www.summarizepaper.com/en/paper/2006.08251v1/

## Key points

- Paper titled "Adversarial Weighting for Domain Adaptation in Regression"
- Authors: Antoine de Mathelin, Guillaume Richard, Mathilde Mougeot, Nicolas Vayatis
- Novel instance-based approach for regression tasks in supervised domain adaptation
- Utilizes innovative optimization objective formulation and adversarial network algorithm
- Adjusts weights of source instances during training to learn task in target domain effectively
- Rigorous experimentation on public datasets
- Provides reproducible results through online demo interface
- Advances regression tasks and offers new possibilities for supervised domain adaptation with improved efficiency and accuracy

## Summary

In their paper titled "Adversarial Weighting for Domain Adaptation in Regression," authors Antoine de Mathelin, Guillaume Richard, Mathilde Mougeot, and Nicolas Vayatis present a novel instance-based approach to address regression tasks within the framework of supervised domain adaptation. Their approach utilizes an innovative formulation of the optimization objective and an adversarial network algorithm to effectively learn the task in the target domain by adjusting weights of source instances during training. Through rigorous experimentation on public datasets and providing reproducible results through an online demo interface, the authors showcase the effectiveness and relevance of their proposed method. This groundbreaking research not only advances regression tasks but also offers new possibilities for handling supervised domain adaptation with improved efficiency and accuracy.