# GAN-Based Synthetic Data

> Source: https://sukruyusufkaya.com/en/glossary/gan-tabanli-sentetik-veri
> Updated: 2026-05-13T21:11:24.661Z
> Type: glossary
> Category: veri-bilimi-ve-veri-yonetimi
**TLDR:** A synthetic data approach based on generating new data samples similar to the real distribution using generative adversarial networks.

<p>GAN-based synthetic data generation is used to create realistic samples, especially in image, tabular, and some sequential data settings. Through the competition between a generator and a discriminator, complex structures of the data distribution can be learned. This approach is attractive for data augmentation, test-case generation, and privacy-oriented data sharing. However, risks such as mode collapse, training instability, and generating samples too close to real data must be managed carefully. Strong generation and safe usage must be considered together.</p>