Filters








5,524 Hits in 4.1 sec

Leveraging the Exact Likelihood of Deep Latent Variable Models [article]

Pierre-Alexandre Mattei, Jes Frellsen
2018 arXiv   pre-print
Finally, we describe an algorithm for missing data imputation using the exact conditional likelihood of a deep latent variable model.  ...  Deep latent variable models (DLVMs) combine the approximation abilities of deep neural networks and the statistical foundations of generative models.  ...  On the boundedness of the likelihood of deep latent variable models Deep generative models with Gaussian outputs assume that the data space is X = R p , and that the family of output distributions is the  ... 
arXiv:1802.04826v4 fatcat:skj22sajfndahftqhxaamrqcpy

Unsupervised Source Separation via Bayesian Inference in the Latent Domain [article]

Michele Mancusi, Emilian Postolache, Giorgio Mariani, Marco Fumero, Andrea Santilli, Luca Cosmo, Emanuele Rodolà
2022 arXiv   pre-print
We leverage the low cardinality of the discrete latent space, trained with a novel loss term imposing a precise arithmetic structure on it, to perform exact Bayesian inference without relying on an approximation  ...  Our algorithm relies on deep Bayesian priors in the form of pre-trained autoregressive networks to model the probability distributions of each source.  ...  Latent likelihood via LQ-VAE In this section we describe how we model the likelihood function and introduce the LQ-VAE model.  ... 
arXiv:2110.05313v4 fatcat:xqymxfib2fhbhayvou6v7rujou

Density estimation using Real NVP [article]

Laurent Dinh, Jascha Sohl-Dickstein, Samy Bengio
2017 arXiv   pre-print
algorithm with exact log-likelihood computation, exact sampling, exact inference of latent variables, and an interpretable latent space.  ...  We demonstrate its ability to model natural images on four datasets through sampling, log-likelihood evaluation and latent variable manipulations.  ...  Acknowledgments The authors thank the developers of Tensorflow [1] . We thank Sherry Moore, David Andersen and Jon Shlens for their help in implementing the model.  ... 
arXiv:1605.08803v3 fatcat:qwjme7s4vvhrzjvvx3d3uu7zqi

Hybrid Models with Deep and Invertible Features [article]

Eric Nalisnick, Akihiro Matsukawa, Yee Whye Teh, Dilan Gorur, Balaji Lakshminarayanan
2019 arXiv   pre-print
The availability of the exact joint density p(targets, features) also allows us to compute many quantities readily, making our hybrid model a useful building block for downstream applications of probabilistic  ...  We propose a neural hybrid model consisting of a linear model defined on a set of features computed by a deep, invertible transformation (i.e. a normalizing flow).  ...  These models are defined by composing invertible functions, and therefore the change-of-variables formula can be used to compute exact densities.  ... 
arXiv:1902.02767v2 fatcat:oh57etkxuzfkbf7zlk4wdeyroa

Stochastic Variational Deep Kernel Learning [article]

Andrew Gordon Wilson, Zhiting Hu, Ruslan Salakhutdinov, Eric P. Xing
2016 arXiv   pre-print
Deep kernel learning combines the non-parametric flexibility of kernel methods with the inductive biases of deep learning architectures.  ...  Specifically, we apply additive base kernels to subsets of output features from deep neural architectures, and jointly learn the parameters of the base kernels and deep network through a Gaussian process  ...  as well as the distributed GP latent variable model (denoted as D-GPLVM) [9] .  ... 
arXiv:1611.00336v2 fatcat:tdi46dwdejd3teezh3gkoqdjhm

The frontier of simulation-based inference [article]

Kyle Cranmer, Johann Brehmer, Gilles Louppe
2020 arXiv   pre-print
We review the rapidly developing field of simulation-based inference and identify the forces giving new momentum to the field.  ...  Many domains of science have developed complex simulations to describe phenomena of interest.  ...  GL is recipient of the ULiège-NRB Chair on Big Data and is thankful for the support of NRB.  ... 
arXiv:1911.01429v3 fatcat:kv32pqap5ne2hkvnekcck4hxkq

SIReN-VAE: Leveraging Flows and Amortized Inference for Bayesian Networks [article]

Jacobie Mouton, Steve Kroon
2022 arXiv   pre-print
a richer class of distributions for the approximate posterior, and stacking layers of latent variables allows more complex priors to be specified for the generative model.  ...  Initial work on variational autoencoders assumed independent latent variables with simple distributions.  ...  ., 2014) provide a powerful framework for constructing deep latent variable models.  ... 
arXiv:2204.11847v1 fatcat:53jbdpa7ovaolc555pehqgzdmu

UNITE: Uncertainty-based Health Risk Prediction Leveraging Multi-sourced Data [article]

Chacha Chen, Junjie Liang, Fenglong Ma, Lucas M. Glass, Jimeng Sun, Cao Xiao
2021 arXiv   pre-print
To achieve model reliability, the model needs to provide accurate prediction and uncertainty score of the prediction.  ...  Successful health risk prediction demands accuracy and reliability of the model.  ...  (II) Stochastic Variational Inference Module Exact inference and learning in Gaussian process model with a non-Gaussian likelihood is not analytically tractable.  ... 
arXiv:2010.11389v2 fatcat:ftmbrzotvzgnldjb3afsieznfq

The frontier of simulation-based inference

Kyle Cranmer, Johann Brehmer, Gilles Louppe
2020 Proceedings of the National Academy of Sciences of the United States of America  
We review the rapidly developing field of simulation-based inference and identify the forces giving additional momentum to the field.  ...  Many domains of science have developed complex simulations to describe phenomena of interest.  ...  modeled with stochastic simulations with billions of latent variables.  ... 
doi:10.1073/pnas.1912789117 pmid:32471948 fatcat:2dabtkqwtzf6ngy62naz3tvlpy

The shape variational autoencoder: A deep generative model of part-segmented 3D objects

C. Nash, C. K. I. Williams
2017 Computer graphics forum (Print)  
Our model makes use of a deep encoder-decoder architecture that leverages the partdecomposability of 3D objects to embed high-dimensional shape representations and sample novel instances.  ...  We provide a quantitative evaluation of the ShapeVAE on shape-completion and test-set log-likelihood tasks and demonstrate that the model performs favourably against strong baselines.  ...  We take a powerful class of deep generative model, the variational autoencoder, and introduce a novel architecture that leverages the hierarchichal part-structure of 3D objects.  ... 
doi:10.1111/cgf.13240 fatcat:7hstjkzgnzdnfo7l6tbge7x32e

Why Normalizing Flows Fail to Detect Out-of-Distribution Data [article]

Polina Kirichenko, Pavel Izmailov, Andrew Gordon Wilson
2020 arXiv   pre-print
Normalizing flows are flexible deep generative models that often surprisingly fail to distinguish between in- and out-of-distribution data: a flow trained on pictures of clothing assigns higher likelihood  ...  We show that by modifying the architecture of flow coupling layers we can bias the flow towards learning the semantic structure of the target data, improving OOD detection.  ...  Introduction Normalizing flows [39, 9, 10] seem to be ideal candidates for out-of-distribution detection, since they are simple generative models that provide an exact likelihood.  ... 
arXiv:2006.08545v1 fatcat:7etzvijmwffjpjpgf3h2bdbb7y

Self-Reflective Variational Autoencoder [article]

Ifigeneia Apostolopoulou, Elan Rosenfeld, Artur Dubrawski
2020 arXiv   pre-print
The Variational Autoencoder (VAE) is a powerful framework for learning probabilistic latent variable generative models.  ...  By redesigning the hierarchical structure of existing VAE architectures, self-reflection ensures that the stochastic flow preserves the factorization of the exact posterior, sequentially updating the latent  ...  Taking a different approach, hierarchical VAEs [2, 7, 12, 13, 14] leverage increasingly deep and interdependent layers of latent variables, similar to how subsequent layers in a discriminative network  ... 
arXiv:2007.05166v1 fatcat:2tkfb42plbboxdwv2g4jobcunq

Efficient Deep Gaussian Process Models for Variable-Sized Input [article]

Issam H. Laradji, Mark Schmidt, Vladimir Pavlovic, Minyoung Kim
2019 arXiv   pre-print
The key advantage is that the combination of GP and DRF leads to a tractable model that can both handle a variable-sized input as well as learn deep long-range dependency structures of the data.  ...  Deep Gaussian processes (DGP) have appealing Bayesian properties, can handle variable-sized data, and learn deep features. Their limitation is that they do not scale well with the size of the data.  ...  ACKNOWLEDGEMENTS We would like to thank the anonymous referees for their constructive comments and suggestions. Issam Laradji was funded by the UBC Four-Year Doctoral Fellowships (4YF).  ... 
arXiv:1905.06982v1 fatcat:xoiukd3tbrgwbjar5gkmqeltey

Deep Probabilistic Programming [article]

Dustin Tran, Matthew D. Hoffman, Rif A. Saurous, Eugene Brevdo, Kevin Murphy, David M. Blei
2017 arXiv   pre-print
In addition, Edward can reuse the modeling representation as part of inference, facilitating the design of rich variational models and generative adversarial networks.  ...  For flexibility, Edward makes it easy to fit the same model using a variety of composable inference methods, ranging from point estimation to variational inference to MCMC.  ...  ACKNOWLEDGEMENTS We thank the probabilistic programming community-for sharing our enthusiasm and motivating further work-including developers of Church, Venture, Gamalon, Hakaru, and WebPPL.  ... 
arXiv:1701.03757v2 fatcat:f3zxlird3bbpblw2fcrq3zuypm

A Meta Learning Approach to Discerning Causal Graph Structure [article]

Justin Wong, Dominik Damjakob
2021 arXiv   pre-print
By interpreting the model predictions as stochastic events, we propose a simple ensemble method classifier to reduce the outcome variability as an average of biased events.  ...  We explore the usage of meta-learning to derive the causal direction between variables by optimizing over a measure of distribution simplicity.  ...  The approach by Goudat et al. (2018) is a try at a more general approach through leverage a series of generative models in order to model each of the observable states of the graph.  ... 
arXiv:2106.05859v1 fatcat:z7lzw3qzy5apbch36zkhdqh55e
« Previous Showing results 1 — 15 out of 5,524 results