{"url":"/method/feature-matching","slug":"feature-matching","name":"GAN Feature Matching","full_name":"GAN Feature Matching","full_name_withheld":false,"description_markdown":"**Feature Matching** is a regularizing objective for a generator in [generative adversarial networks](https://paperswithcode.com/methods/category/generative-adversarial-networks) that prevents it from overtraining on the current discriminator. Instead of directly maximizing the output of the discriminator, the new objective requires the generator to generate data that matches the statistics of the real data, where we use the discriminator only to specify the statistics that we think are worth matching. Specifically, we train the generator to match the expected value of the features on an intermediate layer of the discriminator. This is a natural choice of statistics for the generator to match, since by training the discriminator we ask it to find those features that are most discriminative of real data versus data generated by the current model.\r\n\r\nLetting $\\mathbf{f}\\left(\\mathbf{x}\\right)$ denote activations on an intermediate layer of the discriminator, our new objective for the generator is defined as: $ ||\\mathbb{E}\\_{x\\sim p\\_{data} } \\mathbf{f}\\left(\\mathbf{x}\\right) − \\mathbb{E}\\_{\\mathbf{z}∼p\\_{\\mathbf{z}}\\left(\\mathbf{z}\\right)}\\mathbf{f}\\left(G\\left(\\mathbf{z}\\right)\\right)||^{2}\\_{2} $. The discriminator, and hence\r\n$\\mathbf{f}\\left(\\mathbf{x}\\right)$, are trained as with vanilla GANs. As with regular [GAN](https://paperswithcode.com/method/gan) training, the objective has a fixed point where G exactly matches the distribution of training data.","description_state":"present","introduced_year":null,"introduced_by":{"title":"Improved Techniques for Training GANs","paper":"/paper/improved-techniques-for-training-gans","first_author":"Tim Salimans","n_authors":6,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/improved-techniques-for-training-gans"},"source":{"url":"http://arxiv.org/abs/1606.03498v1","title":"Improved Techniques for Training GANs","url_on_a_paper_host":true},"code_snippet_url":"https://github.com/eli5168/improved_gan_pytorch/blob/f93520646f150a19084efd65215b6c082d3e9d20/improved_GAN.py#L248","code_snippet_url_on_a_code_host":true,"categories":[{"area":"General","area_id":"general","collection":"Regularization","url":"/methods/category/regularization","pwc_aliases":[]}],"n_papers_tagged":16,"archive_num_papers":16,"papers_newest_first":[{"paper":"/paper/stock-market-forecasting-using-dragan-and","title":"Stock market forecasting using DRAGAN and feature matching","date":"2023-01-06","arxiv_id":"2301.05693","n_code_links":1,"syntology":null},{"paper":"/paper/vocbench-a-neural-vocoder-benchmark-for","title":"VocBench: A Neural Vocoder Benchmark for Speech Synthesis","date":"2021-12-06","arxiv_id":"2112.03099","n_code_links":1,"syntology":null},{"paper":"/paper/feature-quantization-improves-gan-training","title":"Feature Quantization Improves GAN Training","date":"2020-04-05","arxiv_id":"2004.02088","n_code_links":2,"syntology":{"ran":1,"of":4,"unverified":3,"pointer_only":1}},{"paper":null,"title":"Imbalanced Data Learning by Minority Class Augmentation using Capsule Adversarial Networks","date":"2020-04-05","arxiv_id":"2004.02182","n_code_links":0,"syntology":null},{"paper":null,"title":"Learning the Loss Functions in a Discriminative Space for Video Restoration","date":"2020-03-20","arxiv_id":"2003.09124","n_code_links":0,"syntology":null},{"paper":"/paper/image-fine-grained-inpainting","title":"Image Fine-grained Inpainting","date":"2020-02-07","arxiv_id":"2002.02609","n_code_links":3,"syntology":null},{"paper":null,"title":"Semi-Supervised Self-Growing Generative Adversarial Networks for Image Recognition","date":"2019-08-11","arxiv_id":"1908.03850","n_code_links":0,"syntology":null},{"paper":"/paper/few-shot-3d-multi-modal-medical-image","title":"Few-shot 3D Multi-modal Medical Image Segmentation using Generative Adversarial Learning","date":"2018-10-29","arxiv_id":"1810.12241","n_code_links":1,"syntology":{"ran":0,"of":6,"unverified":6,"pointer_only":0}},{"paper":null,"title":"A Generative Model of Textures Using Hierarchical Probabilistic Principal Component Analysis","date":"2018-10-16","arxiv_id":"1810.06892","n_code_links":0,"syntology":null},{"paper":null,"title":"Out-of-domain Detection based on Generative Adversarial Network","date":"2018-10-01","arxiv_id":null,"n_code_links":0,"syntology":null},{"paper":null,"title":"Adversarial Training for Adverse Conditions: Robust Metric Localisation using Appearance Transfer","date":"2018-03-09","arxiv_id":"1803.03341","n_code_links":0,"syntology":null},{"paper":null,"title":"Novelty Detection with GAN","date":"2018-02-28","arxiv_id":"1802.10560","n_code_links":0,"syntology":null},{"paper":"/paper/bayesian-gan-1","title":"Bayesian GAN","date":"2017-12-01","arxiv_id":null,"n_code_links":1,"syntology":null},{"paper":"/paper/bayesian-gan","title":"Bayesian GAN","date":"2017-05-26","arxiv_id":"1705.09558","n_code_links":4,"syntology":null},{"paper":"/paper/cvae-gan-fine-grained-image-generation","title":"CVAE-GAN: Fine-Grained Image Generation through Asymmetric Training","date":"2017-03-29","arxiv_id":"1703.10155","n_code_links":3,"syntology":null},{"paper":"/paper/improved-techniques-for-training-gans","title":"Improved Techniques for Training GANs","date":"2016-06-10","arxiv_id":"1606.03498","n_code_links":46,"syntology":{"ran":1,"of":2,"unverified":1,"pointer_only":0}}],"papers_shown":16,"tasks":[{"task":null,"name":"Generative Adversarial Network","papers":3},{"task":"/task/image-generation","name":"Image Generation","papers":3},{"task":"/task/attribute","name":"Attribute","papers":2},{"task":"/task/conditional-image-generation","name":"Conditional Image Generation","papers":2},{"task":"/task/image-inpainting","name":"Image Inpainting","papers":2},{"task":"/task/3d-medical-imaging-segmentation","name":"3D Medical Imaging Segmentation","papers":1},{"task":"/task/appearance-transfer","name":"Appearance Transfer","papers":1},{"task":"/task/brain-image-segmentation","name":"Brain Image Segmentation","papers":1},{"task":"/task/brain-segmentation","name":"Brain Segmentation","papers":1},{"task":"/task/data-augmentation","name":"Data Augmentation","papers":1},{"task":"/task/deblurring","name":"Deblurring","papers":1},{"task":"/task/dimensionality-reduction","name":"Dimensionality Reduction","papers":1},{"task":"/task/face-generation","name":"Face Generation","papers":1},{"task":"/task/face-recognition","name":"Face Recognition","papers":1},{"task":"/task/facial-inpainting","name":"Facial Inpainting","papers":1},{"task":"/task/few-shot-image-segmentation","name":"Few-Shot Semantic Segmentation","papers":1},{"task":"/task/fine-grained-image-inpainting","name":"Fine-Grained Image Inpainting","papers":1},{"task":"/task/classification","name":"General Classification","papers":1},{"task":"/task/image-segmentation","name":"Image Segmentation","papers":1},{"task":"/task/image-to-image-translation","name":"Image-to-Image Translation","papers":1}],"tasks_shown":20,"n_tasks":35,"usage_by_year":[{"year":"2016","papers":1},{"year":"2017","papers":3},{"year":"2018","papers":5},{"year":"2019","papers":1},{"year":"2020","papers":4},{"year":"2021","papers":1},{"year":"2023","papers":1}],"row_source":"methods_table","archive":{"source":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","archive_url":"https://paperswithcode.com/method/feature-matching"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}