{"url":"/method/normalizing-flows","slug":"normalizing-flows","name":"Normalizing Flows","full_name":"Normalizing Flows","full_name_withheld":false,"description_markdown":"**Normalizing Flows** are a method for constructing complex distributions by transforming a\r\nprobability density through a series of invertible mappings. By repeatedly applying the rule for change of variables, the initial density ‘flows’ through the sequence of invertible mappings. At the end of this sequence we obtain a valid probability distribution and hence this type of flow is referred to as a normalizing flow.\r\n\r\nIn the case of finite flows, the basic rule for the transformation of densities considers an invertible, smooth mapping $f : \\mathbb{R}^{d} \\rightarrow \\mathbb{R}^{d}$ with inverse $f^{-1} = g$, i.e. the composition $g \\cdot f\\left(z\\right) = z$. If we use this mapping to transform a random variable $z$ with distribution $q\\left(z\\right)$, the resulting random variable $z' = f\\left(z\\right)$ has a distribution:\r\n\r\n$$ q\\left(\\mathbf{z}'\\right) = q\\left(\\mathbf{z}\\right)\\bigl\\vert{\\text{det}}\\frac{\\delta{f}^{-1}}{\\delta{\\mathbf{z'}}}\\bigr\\vert = q\\left(\\mathbf{z}\\right)\\bigl\\vert{\\text{det}}\\frac{\\delta{f}}{\\delta{\\mathbf{z}}}\\bigr\\vert ^{-1} $$\r\n\f\r\nwhere the last equality can be seen by applying the chain rule (inverse function theorem) and is a property of Jacobians of invertible functions. We can construct arbitrarily complex densities by composing several simple maps and successively applying the above equation. The density $q\\_{K}\\left(\\mathbf{z}\\right)$ obtained by successively transforming a random variable $z\\_{0}$ with distribution $q\\_{0}$ through a chain of $K$ transformations $f\\_{k}$ is:\r\n\r\n$$ z\\_{K} = f\\_{K} \\cdot \\dots \\cdot f\\_{2} \\cdot f\\_{1}\\left(z\\_{0}\\right) $$\r\n\r\n$$ \\ln{q}\\_{K}\\left(z\\_{K}\\right) = \\ln{q}\\_{0}\\left(z\\_{0}\\right) − \\sum^{K}\\_{k=1}\\ln\\vert\\det\\frac{\\delta{f\\_{k}}}{\\delta{\\mathbf{z\\_{k-1}}}}\\vert $$\r\n\f\r\nThe path traversed by the random variables $z\\_{k} = f\\_{k}\\left(z\\_{k-1}\\right)$ with initial distribution $q\\_{0}\\left(z\\_{0}\\right)$ is called the flow and the path formed by the successive distributions $q\\_{k}$ is a normalizing flow.","description_state":"present","introduced_year":null,"introduced_by":{"title":"Variational Inference with Normalizing Flows","paper":"/paper/variational-inference-with-normalizing-flows","first_author":"Danilo Jimenez Rezende","n_authors":2,"url_abs":null,"archive_paper_url":"https://paperswithcode.com/paper/variational-inference-with-normalizing-flows"},"source":{"url":"http://arxiv.org/abs/1505.05770v6","title":"Variational Inference with Normalizing Flows","url_on_a_paper_host":true},"code_snippet_url":"https://github.com/ex4sperans/variational-inference-with-normalizing-flows/blob/922b569f851e02fa74700cd0754fe2ef5c1f3180/flow.py#L9","code_snippet_url_on_a_code_host":true,"categories":[{"area":"General","area_id":"general","collection":"Distribution Approximation","url":"/methods/category/distribution-approximation","pwc_aliases":[]}],"n_papers_tagged":771,"archive_num_papers":771,"papers_newest_first":[{"paper":"/paper/latent-thermodynamic-flows-unified","title":"Latent Thermodynamic Flows: Unified Representation Learning and Generative Modeling of Temperature-Dependent Behaviors from Limited Data","date":"2025-07-03","arxiv_id":"2507.03174","n_code_links":1,"syntology":null},{"paper":null,"title":"Distilling Normalizing Flows","date":"2025-06-26","arxiv_id":"2506.21003","n_code_links":0,"syntology":null},{"paper":"/paper/textrm-ode-t-left-textrm-ode-l-right","title":"$\\textrm{ODE}_t \\left(\\textrm{ODE}_l \\right)$: Shortcutting the Time and Length in Diffusion and Flow Models for Faster Sampling","date":"2025-06-26","arxiv_id":"2506.21714","n_code_links":1,"syntology":null},{"paper":null,"title":"Time-series surrogates from energy consumers generated by machine learning approaches for long-term forecasting scenarios","date":"2025-06-25","arxiv_id":"2506.20253","n_code_links":0,"syntology":null},{"paper":"/paper/operator-forces-for-coarse-grained-molecular","title":"Operator Forces For Coarse-Grained Molecular Dynamics","date":"2025-06-24","arxiv_id":"2506.19628","n_code_links":1,"syntology":null},{"paper":null,"title":"Simulating Correlated Electrons with Symmetry-Enforced Normalizing Flows","date":"2025-06-20","arxiv_id":"2506.17015","n_code_links":0,"syntology":null},{"paper":null,"title":"Latent Noise Injection for Private and Statistically Aligned Synthetic Data Generation","date":"2025-06-19","arxiv_id":"2506.16636","n_code_links":0,"syntology":null},{"paper":null,"title":"Characterizing Neural Manifolds' Properties and Curvatures using Normalizing Flows","date":"2025-06-13","arxiv_id":"2506.12187","n_code_links":0,"syntology":null},{"paper":"/paper/simulation-trained-conditional-normalizing","title":"Simulation-trained conditional normalizing flows for likelihood approximation: a case study in stress regulation kinetics in yeast","date":"2025-06-11","arxiv_id":"2506.09374","n_code_links":1,"syntology":null},{"paper":"/paper/emulating-compact-binary-population-synthesis","title":"Emulating compact binary population synthesis simulations with robust uncertainty quantification and model comparison: Bayesian normalizing flows","date":"2025-06-06","arxiv_id":"2506.05657","n_code_links":1,"syntology":null},{"paper":"/paper/2506-04749","title":"Amortized variational transdimensional inference","date":"2025-06-05","arxiv_id":"2506.04749","n_code_links":1,"syntology":null},{"paper":null,"title":"MARBLE: Material Recomposition and Blending in CLIP-Space","date":"2025-06-05","arxiv_id":"2506.05313","n_code_links":0,"syntology":null},{"paper":"/paper/savage-dickey-density-ratio-estimation-with","title":"Savage-Dickey density ratio estimation with normalizing flows for Bayesian model comparison","date":"2025-06-04","arxiv_id":"2506.04339","n_code_links":1,"syntology":null},{"paper":null,"title":"Asymptotically exact variational flows via involutive MCMC kernels","date":"2025-06-02","arxiv_id":"2506.02162","n_code_links":0,"syntology":null},{"paper":"/paper/promptvfx-text-driven-fields-for-open-world","title":"PromptVFX: Text-Driven Fields for Open-World 3D Gaussian Animation","date":"2025-06-01","arxiv_id":"2506.01091","n_code_links":1,"syntology":null},{"paper":null,"title":"Quality Assessment of Noisy and Enhanced Speech with Limited Data: UWB-NTIS System for VoiceMOS 2024 and Beyond","date":"2025-05-31","arxiv_id":"2506.00506","n_code_links":0,"syntology":null},{"paper":null,"title":"Learning geometry and topology via multi-chart flows","date":"2025-05-30","arxiv_id":"2505.24665","n_code_links":0,"syntology":null},{"paper":null,"title":"DiCoFlex: Model-agnostic diverse counterfactuals with flexible control","date":"2025-05-29","arxiv_id":"2505.23700","n_code_links":0,"syntology":null},{"paper":"/paper/normalizing-flows-are-capable-models-for-rl","title":"Normalizing Flows are Capable Models for RL","date":"2025-05-29","arxiv_id":"2505.23527","n_code_links":1,"syntology":null},{"paper":null,"title":"SESaMo: Symmetry-Enforcing Stochastic Modulation for Normalizing Flows","date":"2025-05-26","arxiv_id":"2505.19619","n_code_links":0,"syntology":null},{"paper":"/paper/hybrid-bernstein-normalizing-flows-for","title":"Hybrid Bernstein Normalizing Flows for Flexible Multivariate Density Regression with Interpretable Marginals","date":"2025-05-20","arxiv_id":"2505.14164","n_code_links":2,"syntology":null},{"paper":"/paper/flowpure-continuous-normalizing-flows-for","title":"FlowPure: Continuous Normalizing Flows for Adversarial Purification","date":"2025-05-19","arxiv_id":"2505.13280","n_code_links":1,"syntology":null},{"paper":null,"title":"Minimum-Excess-Work Guidance","date":"2025-05-19","arxiv_id":"2505.13375","n_code_links":0,"syntology":null},{"paper":"/paper/trustworthy-image-super-resolution-via","title":"Trustworthy Image Super-Resolution via Generative Pseudoinverse","date":"2025-05-18","arxiv_id":"2505.12375","n_code_links":1,"syntology":null},{"paper":null,"title":"Path Gradients after Flow Matching","date":"2025-05-15","arxiv_id":"2505.10139","n_code_links":0,"syntology":null},{"paper":null,"title":"Continuous Temporal Learning of Probability Distributions via Neural ODEs with Applications in Continuous Glucose Monitoring Data","date":"2025-05-13","arxiv_id":"2505.08698","n_code_links":0,"syntology":null},{"paper":null,"title":"Contrastive Normalizing Flows for Uncertainty-Aware Parameter Estimation","date":"2025-05-13","arxiv_id":"2505.08709","n_code_links":0,"syntology":null},{"paper":null,"title":"Hamiltonian Normalizing Flows as kinetic PDE solvers: application to the 1D Vlasov-Poisson Equations","date":"2025-05-07","arxiv_id":"2505.04471","n_code_links":0,"syntology":null},{"paper":"/paper/physics-informed-sylvester-normalizing-flows","title":"Physics-Informed Sylvester Normalizing Flows for Bayesian Inference in Magnetic Resonance Spectroscopy","date":"2025-05-06","arxiv_id":"2505.03590","n_code_links":1,"syntology":null},{"paper":null,"title":"Continuous Normalizing Flows for Uncertainty-Aware Human Pose Estimation","date":"2025-05-04","arxiv_id":"2505.02287","n_code_links":0,"syntology":null}],"papers_shown":30,"tasks":[{"task":"/task/density-estimation","name":"Density Estimation","papers":76},{"task":"/task/variational-inference","name":"Variational Inference","papers":54},{"task":"/task/image-generation","name":"Image Generation","papers":37},{"task":"/task/anomaly-detection","name":"Anomaly Detection","papers":29},{"task":"/task/text-to-speech","name":"Text to Speech","papers":26},{"task":"/task/text-to-speech-1","name":"text-to-speech","papers":26},{"task":"/task/uncertainty-quantification","name":"Uncertainty Quantification","papers":24},{"task":"/task/time-series-1","name":"Time Series","papers":23},{"task":"/task/speech-synthesis","name":"Speech Synthesis","papers":22},{"task":"/task/bayesian-inference","name":"Bayesian Inference","papers":18},{"task":"/task/denoising","name":"Denoising","papers":18},{"task":"/task/regression-1","name":"regression","papers":17},{"task":"/task/out-of-distribution-detection","name":"Out-of-Distribution Detection","papers":16},{"task":"/task/time-series","name":"Time Series Analysis","papers":16},{"task":"/task/decision-making","name":"Decision Making","papers":15},{"task":"/task/gaussian-processes","name":"Gaussian Processes","papers":15},{"task":"/task/reinforcement-learning-1","name":"Reinforcement Learning (RL)","papers":15},{"task":"/task/decoder","name":"Decoder","papers":14},{"task":"/task/representation-learning","name":"Representation Learning","papers":14},{"task":"/task/reinforcement-learning-2","name":"reinforcement-learning","papers":14}],"tasks_shown":20,"n_tasks":367,"usage_by_year":[{"year":"2015","papers":7},{"year":"2016","papers":13},{"year":"2017","papers":25},{"year":"2018","papers":30},{"year":"2019","papers":69},{"year":"2020","papers":100},{"year":"2021","papers":115},{"year":"2022","papers":106},{"year":"2023","papers":139},{"year":"2024","papers":110},{"year":"2025","papers":57}],"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/normalizing-flows"},"syntology_read_at":"2026-09-24T18:15:14+00:00"}