# GAN-supervised Seismic Data Reconstruction: An Enhanced-Learning for Improved Generalization

**Authors:** Paul Goyes-Penafiel, Leon Suarez-Rodriguez, Claudia Correa, Henry Arguello  
**arXiv:** [2311.10910](https://arxiv.org/abs/2311.10910)  
**Date:** 2023-11-17  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2311.10910v2/

## Key points

- Title: "GAN-supervised Seismic Data Reconstruction: An Enhanced-Learning for Improved Generalization"
- Authors: Paul Goyes-Penafiel, Leon Suarez-Rodriguez, Claudia Correa, Henry Arguello
- Challenges addressed: Seismic data interpolation of irregularly missing traces
- Importance of process: Crucial for subsurface imaging for accurate analysis and interpretation
- Proposed approach: Novel deep-learning-based method leveraging data redundancy
- Training process: Two-stage training using GAN and reconstruction network
- Outcome: Demonstrated effectiveness in improving overall performance on field and synthetic seismic datasets

## Summary

, , , , 

The paper titled "GAN-supervised Seismic Data Reconstruction: An Enhanced-Learning for Improved Generalization" by Paul Goyes-Penafiel, Leon Suarez-Rodriguez, Claudia Correa, and Henry Arguello addresses the challenges in seismic data interpolation of irregularly missing traces. This process is crucial for subsurface imaging as it enables accurate analysis and interpretation throughout the seismic processing workflow. The authors propose a novel deep-learning-based approach that leverages data redundancy to overcome limitations in traditional methods. The method involves a two-stage training process using an adversarial generative network (GAN) and a reconstruction network to enhance generalization capabilities. Experimental results on both field and synthetic seismic datasets demonstrate the effectiveness of this approach in improving overall performance.