How deep are deep Gaussian processes?
From MaRDI portal
Abstract: Recent research has shown the potential utility of Deep Gaussian Processes. These deep structures are probability distributions, designed through hierarchical construction, which are conditionally Gaussian. In this paper, the current published body of work is placed in a common framework and, through recursion, several classes of deep Gaussian processes are defined. The resulting samples generated from a deep Gaussian process have a Markovian structure with respect to the depth parameter, and the effective depth of the resulting process is interpreted in terms of the ergodicity, or non-ergodicity, of the resulting Markov chain. For the classes of deep Gaussian processes introduced, we provide results concerning their ergodicity and hence their effective depth. We also demonstrate how these processes may be used for inference; in particular we show how a Metropolis-within-Gibbs construction across the levels of the hierarchy can be used to derive sampling tools which are robust to the level of resolution used to represent the functions on a computer. For illustration, we consider the effect of ergodicity in some simple numerical examples.
Recommendations
Cites work
- A Bayesian level set method for geometric inverse problems
- A general framework for the parametrization of hierarchical models
- An explicit link between Gaussian fields and Gaussian Markov random fields: the stochastic partial differential equation approach
- Analysis of SPDEs arising in path sampling. I: The Gaussian case
- Bayesian calibration of computer models. (With discussion)
- Bayesian Inference for Non-Stationary Spatial Covariance Structure via Spatial Deformations
- Bayesian learning for neural networks
- Combining Field Data and Computer Simulations for Calibration and Prediction
- Ergodicity for SDEs and approximations: locally Lipschitz vector fields and degenerate noise.
- Gaussian processes for machine learning.
- scientific article; zbMATH DE number 3718234 (Why is no real title available?)
- scientific article; zbMATH DE number 1405266 (Why is no real title available?)
- Infinite-dimensional dynamical systems. An introduction to dissipative parabolic PDEs and the theory of global attractors
- Interpolation of spatial data. Some theory for kriging
- Iterated Random Functions
- Kullback-Leibler approximation for probability measures on infinite dimensional spaces
- Markov Chains and Stochastic Stability
- MCMC methods for functions: modifying old algorithms to make them faster
- Mitigating the influence of the boundary on PDE-based covariance operators
- Parameterizations for ensemble Kalman inversion
- Scattered Data Approximation
- Stable architectures for deep neural networks
- Statistical and computational inverse problems.
- Whittle-Matérn priors for Bayesian statistical inversion with applications in electrical impedance tomography
Cited in
(38)- Emulating dynamic non-linear simulators using Gaussian processes
- Deep state-space Gaussian processes
- Uncertainty quantification of a computer model for binary black hole formation
- Cauchy Markov random field priors for Bayesian inversion
- Likelihood-free inference with deep Gaussian processes
- Posterior inference for sparse hierarchical non-stationary models
- Rejoinder: ``Probabilistic integration: a role in statistical computation?
- A survey of deep network techniques all classifiers can adopt
- Foundations of Bayesian inference for complex statistical models. Abstracts from the workshop held May 2--8, 2021 (hybrid meeting)
- Diagnostics-driven nonstationary emulators using kernel mixtures
- On the well-posedness of Bayesian inverse problems
- Multilevel quasi-Monte Carlo integration with product weights for elliptic PDEs with lognormal coefficients
- Wide neural networks with bottlenecks are deep Gaussian processes
- Non-stationary multi-layered Gaussian priors for Bayesian inversion
- The ridgelet prior: a covariance function approach to prior specification for Bayesian neural networks
- Priors in Bayesian Deep Learning: A Review
- Bayesian inversion with α-stable priors
- Nonstationary Gaussian Process Discriminant Analysis With Variable Selection for High-Dimensional Functional Data
- Deep Compositional Spatial Models
- Bayesian Inverse Problems Are Usually Well-Posed
- Almost surely safe exploration and exploitation for deep reinforcement learning with state safety estimation
- Bayesian emulation of complex computer models with structured partial discontinuities
- Active Learning for Deep Gaussian Process Surrogates
- Hyperpriors for Matérn fields with applications in Bayesian inversion
- ProSpar-GP: Scalable Gaussian Process Modeling with Massive Nonstationary Datasets
- Active learning via heteroskedastic rational kriging
- Convergence rates of non-stationary and deep Gaussian process regression
- Deep latent force models: ODE-based process convolutions for Bayesian deep learning
- Contour location for reliability in airfoil simulation experiments using deep Gaussian processes
- Deep Gaussian process priors for Bayesian image reconstruction
- Kernel-based regularized learning with random projections: beyond least squares
- Gaussian measures conditioned on nonlinear observations: consistency, MAP estimators, and simulation
- Posterior contraction rates for constrained deep Gaussian processes in density estimation and classification
- A primer on Bayesian neural networks: review and debates
- Relaxed Gaussian process interpolation: a goal-oriented approach to Bayesian optimization
- Bayesian emulation for computer models with multiple partial discontinuities
- Vecchia-approximated Deep Gaussian Processes for Computer Experiments
- Deep Gaussian mixture models
This page was built for publication: How deep are deep Gaussian processes?
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4558207)
