Filters








1,375 Hits in 3.2 sec

Density estimation using Real NVP [article]

Laurent Dinh, Jascha Sohl-Dickstein, Samy Bengio
2017 arXiv   pre-print
We extend the space of such models using real-valued non-volume preserving (real NVP) transformations, a set of powerful invertible and learnable transformations, resulting in an unsupervised learning  ...  This model can perform efficient and exact inference, sampling and log-density estimation of data points.  ...  We address this challenge by introducing real-valued non-volume preserving (real NVP) transformations, a tractable yet expressive approach to modeling high-dimensional data.  ... 
arXiv:1605.08803v3 fatcat:qwjme7s4vvhrzjvvx3d3uu7zqi

Masked Autoregressive Flow for Density Estimation [article]

George Papamakarios, Theo Pavlakou, Iain Murray
2018 arXiv   pre-print
This type of flow is closely related to Inverse Autoregressive Flow and is a generalization of Real NVP.  ...  Autoregressive models are among the best performing neural density estimators.  ...  Acknowledgments We thank Maria Gorinova for useful comments, and Johann Brehmer for discovering the error in the calculation of the test log likelihood for conditional MAF.  ... 
arXiv:1705.07057v4 fatcat:svp6kx4ebjdxdpp7zhtfrlrj54

Embarrassingly parallel MCMC using deep invertible transformations [article]

Diego Mesquita, Paul Blomstedt, Samuel Kaski
2021 arXiv   pre-print
Moreover, they enable us to sample from the product of the subposteriors using an efficient and stable importance sampling scheme.  ...  REAL NVP DENSITY ESTIMATION Real NVP (Dinh et al., 2017) is a class of deep generative models in which a D-dimensional real-valued quantity of interest x is modeled as a composition of bijective transformations  ...  In this work, we make use of the fact that bijective transformations using real NVP offers both accurate density estimation and computationally efficient sampling for arbitrarily complex distributions.  ... 
arXiv:1903.04556v2 fatcat:g2fkgzx7anez5d5dfp42v2fyyu

Constrained Density Matching and Modeling for Cross-lingual Alignment of Contextualized Representations [article]

Wei Zhao, Steffen Eger
2022 arXiv   pre-print
To address these issues, we introduce supervised and unsupervised density-based approaches named Real-NVP and GAN-Real-NVP, driven by Normalizing Flow, to perform alignment, both dissecting the alignment  ...  of multilingual subspaces into density matching and density modeling.  ...  To better leverage data, we include density estimation (modeling) based on Real-NVP in the procedure of adversarial training.  ... 
arXiv:2201.13429v1 fatcat:uyicv3l6i5gm3df7c7zv5ndzym

Normalizing flows for random fields in cosmology [article]

Adam Rouhiainen, Utkarsh Giri, Moritz Münchmeyer
2021 arXiv   pre-print
We evaluate the performance of different normalizing flows for both density estimation and sampling of near-Gaussian random fields, and check the quality of samples with different statistics such as power  ...  We explore aspects of these flows that are specific to cosmology, such as flowing from a physical prior distribution and evaluating the density estimation results in the analytically tractable correlated  ...  Basic real NVP flow The first flow with success in creating high quality images was the real-valued non-volume preserving (real NVP ) flow [3] .  ... 
arXiv:2105.12024v1 fatcat:xwb5gsvhrrhlpjv2x45nw7mw54

Comparison of Maximum Likelihood and GAN-based training of Real NVPs [article]

Ivo Danihelka, Balaji Lakshminarayanan, Benigno Uria, Daan Wierstra, Peter Dayan
2017 arXiv   pre-print
Finally, we use ideas from the one-shot learning literature to develop a novel fast learning critic.  ...  We then compare the generated samples, exact log-probability densities and approximate Wasserstein distances.  ...  Log-probability Density Ratio Evaluation Real NVPs are not limited to the computation of negative log-probability densities of visible variables.  ... 
arXiv:1705.05263v1 fatcat:tjxqumbrybenhgwzknqqqw3f3a

Resampling Base Distributions of Normalizing Flows [article]

Vincent Stimper, Bernhard Schölkopf, José Miguel Hernández-Lobato
2022 arXiv   pre-print
2D densities, density estimation of tabular data, image generation, and modeling Boltzmann distributions.  ...  Furthermore, we develop suitable learning algorithms using both maximizing the log-likelihood and the optimization of the Kullback-Leibler divergence, and apply them to various sample problems, i.e. approximating  ...  Moreover, we train real NVP models with Gaussian and resampled base distributions with the reverse KL divergence using the gradient estimators derived in Theorem 2.  ... 
arXiv:2110.15828v2 fatcat:hjjvt3d3jfenfafipsgnxy6ke4

Learnable Explicit Density for Continuous Latent Space and Variational Inference [article]

Chin-Wei Huang, Ahmed Touati, Laurent Dinh, Michal Drozdzal, Mohammad Havaei, Laurent Charlin, Aaron Courville
2017 arXiv   pre-print
First, we decompose the learning of VAEs into layerwise density estimation, and argue that having a flexible prior is beneficial to both sample generation and inference.  ...  Our analysis results in a unified approach to parameterizing a VAE, without the need to restrict ourselves to use factorial Gaussians in the latent real space.  ...  Acknowledgements We thank NVIDIA for donating a DGX-1 computer used in this work.  ... 
arXiv:1710.02248v1 fatcat:caospxxwezcrfd2h5td7tigooe

Learning to Importance Sample in Primary Sample Space [article]

Quan Zheng, Matthias Zwicker
2018 arXiv   pre-print
We use Real NVP to non-linearly warp primary sample space and obtain desired densities.  ...  During a scene-dependent training phase, we learn to generate samples with a desired density in the primary sample space of the rendering algorithm using maximum likelihood estimation.  ...  We use Real NVP to non-linearly warp primary sample space and obtain desired densities.  ... 
arXiv:1808.07840v1 fatcat:fv5ifxp64javnicmu67jvxp4au

Range-kNN queries with privacy protection in a mobile environment

Zhou Shao, David Taniar, Kiki Maulana Adhinugraha
2015 Pervasive and Mobile Computing  
A Landmark Tree (LT), which indexes all the location information, is used to hide the actual user location in a specific radius.  ...  In order to protect the privacy of users' personal information, we proposed Range-kNN queries, which uses the query range instead of one single query point.  ...  In Fig. 18 , four different objects densities are used. The number of changed NVPs is just for one intersection NVP. It is the same to the number of adjacent NVPs for a given NVP.  ... 
doi:10.1016/j.pmcj.2015.05.004 fatcat:3n3rybxlljhmvmc3a7qwoqstjy

Scaling Up Machine Learning For Quantum Field Theory with Equivariant Continuous Flows [article]

Pim de Haan, Corrado Rainone, Miranda C.N. Cheng, Roberto Bondesan
2021 arXiv   pre-print
In contrast to the deep architectures used so far for this task, our proposal is based on a shallow design and incorporates the symmetries of the problem.  ...  NVP A Real NVP flow uses a stack of coupling layers g : z ∈ R N → φ ∈ R N defined as follows.  ...  In section 4 we show that in our experiments that the density induced by a real NVP normalizing flow trained to match the invariant density of the φ 4 theory does not obtain good invariance properties  ... 
arXiv:2110.02673v2 fatcat:nopaq6ghgfb2lofmc4mxhtvijm

Learning Probabilistic Multi-Modal Actor Models for Vision-Based Robotic Grasping [article]

Mengyuan Yan, Adrian Li, Mrinal Kalakrishnan, Peter Pastor
2019 arXiv   pre-print
We demonstrate on both simulation and real robot that the proposed actor model achieves similar performance compared to the value network using the Cross-Entropy Method (CEM) for inference, on top-down  ...  We propose an alternative method, by directly training a neural density model to approximate the conditional distribution of successful grasp poses from the input images.  ...  estimation [3] , [4] , [12] , neural network models such as Real NVP [3] are able to approximate arbitrary distributions.  ... 
arXiv:1904.07319v1 fatcat:2ri4oxrkffhhhbiwv3q5sdpxmi

Log-Likelihood Ratio Minimizing Flows: Towards Robust and Quantifiable Neural Distribution Alignment [article]

Ben Usman, Avneesh Sud, Nick Dufour, Kate Saenko
2020 arXiv   pre-print
However, the former fails to capture the structure of complex real-world distributions, while the latter is difficult to train and does not provide any universal convergence guarantees or automatic quantitative  ...  (the "default" Real NVP).  ...  We used LRMF with Gaussian, Real NVP, and FFJORD densities P M (x; θ) with affine, NVP, and FFJORD transformations T (x; φ) respectively to align pairs of moonshaped and blob-shaped datasets.  ... 
arXiv:2003.12170v2 fatcat:fmjiiub7gndctg4fqnbjxiko2i

A RAD approach to deep mixture models [article]

Laurent Dinh, Jascha Sohl-Dickstein, Hugo Larochelle, Razvan Pascanu
2020 arXiv   pre-print
Flow based models such as Real NVP are an extremely powerful approach to density estimation.  ...  To address this difficulty, we present a normalizing flow architecture which relies on domain partitioning using locally invertible functions, and possesses both real and discrete valued latent variables  ...  (a) Real NVP on grid Gaussian mixture. (b) Real NVP on ring Gaussian mixture. (c) Real NVP on two moons. (d) Real NVP on two circles. (e) Real NVP on spiral. (f) Real NVP on many moons.  ... 
arXiv:1903.07714v3 fatcat:wqte2sxxvrdundng7id7bkgmui

Roundtrip: A Deep Generative Neural Density Estimator [article]

Qiao Liu, Jiaze Xu, Rui Jiang, Wing Hung Wong
2020 arXiv   pre-print
Unlike previous neural density estimators that put stringent conditions on the transformation from the latent space to the data space, Roundtrip enables the use of much more general mappings.  ...  Density estimation is a fundamental problem in both statistics and machine learning.  ...  In these experiments, we compared Roundtrip to the widely used Gaussian kernel density estimator as well as several neural density estimators, including MADE (Germain et al. 2015) , Real NVP (Dinh, Sohl-Dickstein  ... 
arXiv:2004.09017v4 fatcat:qmdazwyjczfdthopo73wqfehtq
« Previous Showing results 1 — 15 out of 1,375 results