# On the Units of GANs (Extended Abstract)

arXiv: https://arxiv.org/abs/1901.09887 · HTML: https://mlsanity.com/paper/1901.09887/

## Key facts

- Cited by 19 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 29 Jan 2019.

## Built on by (most cited first, 19 of 19)

- [Interpretable Deep Learning: Interpretation, Interpretability, Trustworthiness, and Beyond](https://mlsanity.com/paper/2103.10689/) (2021)
- [The Quest for the Right Mediator: Surveying Mechanistic Interpretability Through the Lens of Causal Mediation Analysis](https://mlsanity.com/paper/2408.01416/) (2024)
- [Towards a fuller understanding of neurons with Clustered Compositional Explanations](https://mlsanity.com/paper/2310.18443/) (2023)
- [From Neural Activations to Concepts: A Survey on Explaining Concepts in Neural Networks](https://mlsanity.com/paper/2310.11884/) (2023)
- [Personalized Face Inpainting with Diffusion Models by Parallel Visual Attention](https://mlsanity.com/paper/2312.03556/) (2023)
- [Navigating Beyond Instructions: Vision-and-Language Navigation in Obstructed Environments](https://mlsanity.com/paper/2407.21452/) (2024)
- [What do End-to-End Speech Models Learn about Speaker, Language and Channel Information? A Layer-wise and Neuron-level Analysis](https://mlsanity.com/paper/2107.00439/) (2021)
- [IA-FaceS: A Bidirectional Method for Semantic Face Editing](https://mlsanity.com/paper/2203.13097/) (2022)
- [Understanding Multi-View Transformers](https://mlsanity.com/paper/2510.24907/) (2025)
- [Extracting Semantic Knowledge from GANs with Unsupervised Learning](https://mlsanity.com/paper/2211.16710/) (2022)
- [FreeCond: Free Lunch in the Input Conditions of Text-Guided Inpainting](https://mlsanity.com/paper/2412.00427/) (2024)
- [An Unsupervised Way to Understand Artifact Generating Internal Units in Generative Neural Networks](https://mlsanity.com/paper/2112.08814/) (2021)
- [Can We Find Neurons that Cause Unrealistic Images in Deep Generative Networks?](https://mlsanity.com/paper/2201.06346/) (2022)
- [BiOcularGAN: Bimodal Synthesis and Annotation of Ocular Images](https://mlsanity.com/paper/2205.01536/) (2022)
- [Applications of Explainable artificial intelligence in Earth system science](https://mlsanity.com/paper/2406.11882/) (2024)
- [A Multi-Level Causal Intervention Framework for Mechanistic Interpretability in Variational Autoencoders](https://mlsanity.com/paper/2505.03530/) (2025)
- [On-the-fly Object Detection using StyleGAN with CLIP Guidance](https://mlsanity.com/paper/2210.16742/) (2022)
- [More comprehensive facial inversion for more effective expression recognition](https://mlsanity.com/paper/2211.13564/) (2022)
- [Demystifying What Code Summarization Models Learned](https://mlsanity.com/paper/2303.02333/) (2023)

## Abstract (arXiv)

Generative Adversarial Networks (GANs) have achieved impressive results for many real-world applications. As an active research topic, many GAN variants have emerged with improvements in sample quality and training stability. However, visualization and understanding of GANs is largely missing. How does a GAN represent our visual world internally? What causes the artifacts in GAN results? How do architectural choices affect GAN learning? Answering such questions could enable us to develop new insights and better models. In this work, we present an analytic framework to visualize and understand GANs at the unit-, object-, and scene-level. We first identify a group of interpretable units that are closely related to concepts with a segmentation-based network dissection method. We quantify the causal effect of interpretable units by measuring the ability of interventions to control objects in the output. Finally, we examine the contextual relationship between these units and their surrounding by inserting the discovered concepts into new images. We show several practical applications enabled by our framework, from comparing internal representations across different layers, models, and datasets, to improving GANs by locating and removing artifact-causing units, to interactively manipulating objects in the scene. We will open source our interactive tools to help researchers and practitioners better understand their models.