Taming Normalizing Flows
Not on the map. Why?
Abstract (arXiv)
We propose an algorithm for taming Normalizing Flow models - changing the probability that the model will produce a specific image or image category. We focus on Normalizing Flows because they can calculate the exact generation probability likelihood for a given image. We demonstrate taming using models that generate human faces, a subdomain with many interesting privacy and bias considerations. Our method can be used in the context of privacy, e.g., removing a specific person from the output of a model, and also in the context of debiasing by forcing a model to output specific image categories according to a given target distribution. Taming is achieved with a fast fine-tuning process without retraining the model from scratch, achieving the goal in a matter of minutes. We evaluate our method qualitatively and quantitatively, showing that the generation quality remains intact, while the desired changes are applied.
Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Builds on 42 of its 69 references that are in the map. First posted on arXiv 29 Nov 2022.
Builds on
42 of 69 references.
References · 42
- 2015Deep Residual Learning for Image Recognitioncited by 51,809
- 2014Adam: A Method for Stochastic Optimizationcited by 36,960
- 2020Denoising Diffusion Probabilistic Modelscited by 20,851
- 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
- 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
- 2016Overcoming catastrophic forgetting in neural networkscited by 4,718
- 2014Deep Learning Face Attributes in the Wildcited by 3,679
- 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,967
- 2015Variational Inference with Normalizing Flowscited by 2,161
- 2020Extracting Training Data from Large Language Modelscited by 2,144
- 2016Density estimation using Real NVPcited by 2,029
- 2018Glow: Generative Flow with Invertible 1x1 Convolutionscited by 1,539
30 more
- 2014NICE: Non-linear Independent Components Estimationcited by 1,052
- 2019Machine Unlearningcited by 1,020
- 2013An Empirical Investigation of Catastrophic Forgetting in Gradient-Based Neural Networkscited by 711
- 2019Interpreting the Latent Space of GANs for Semantic Face Editingcited by 691
- 2017Masked Autoregressive Flow for Density Estimationcited by 680
- 2018Mitigating Unwanted Biases with Adversarial Learningcited by 663
- 2019Eternal Sunshine of the Spotless Net: Selective Forgetting in Deep Networkscited by 599
- 2018FFJORD: Free-form Continuous Dynamics for Scalable Reversible Generative Modelscited by 584
- 2019Certified Data Removal from Machine Learning Modelscited by 512
- 2015A note on the evaluation of generative modelscited by 291
- 2019Flow++: Improving Flow-Based Generative Models with Variational Dequantization and Architecture Designcited by 258
- 2020When Machine Unlearning Jeopardizes Privacycited by 181
- 2022Reconstructing Training Data from Trained Neural Networkscited by 129
- 2020Fair Attribute Classification through Latent Space De-biasingcited by 125
- 2019Fair Generative Modeling via Weak Supervisioncited by 114
- 2020Rewriting a Deep Generative Modelcited by 105
- 2021GAN-Control: Explicitly Controllable GANscited by 91
- 2019FairFace: Face Attribute Dataset for Balanced Race, Gender, and Agecited by 87
- 2021Graph-Based Deep Learning for Medical Diagnosis and Analysis: Past, Present and Futurecited by 67
- 2022PUMA: Performance Unchanged Model Augmentation for Training Data Removalcited by 56
- 2020Navigating the GAN Parameter Space for Semantic Image Editingcited by 49
- 2021Densely connected normalizing flowscited by 49
- 2020Improving the Fairness of Deep Generative Models without Retrainingcited by 46
- 2020VFlow: More Expressive Generative Flows with Variational Data Augmentationcited by 38
- 2013Deep Generative Stochastic Networks Trainable by Backpropcited by 36
- 2022Repairing Neural Networks by Leaving the Right Past Behindcited by 27
- 2022Generative Visual Prompt: Unifying Distributional Control of Pre-Trained Generative Modelscited by 27
- 2019E.T.-RNN: Applying Deep Learning to Credit Loan Applicationscited by 26
- 2022Rewriting Geometric Rules of a GANcited by 25
- 2022FairStyle: Debiasing StyleGAN2 with Style Channel Manipulationscited by 20
Built on by
Citing papers per year: 2024: 1; 2025: 0; 2026: 2.
3 papers.