Intrinsic Gaussian processes on complex constrained domains

We propose a class of intrinsic Gaussian processes (in-GPs) for interpolation, regression and classification on manifolds with a primary focus on complex constrained domains or irregular shaped spaces arising as subsets or submanifolds of R, R2, R3 and beyond. For example, in-GPs can accommodate spatial domains arising as complex subsets of Euclidean space. in-GPs respect the potentially complex boundary or interior conditions as well as the intrinsic geometry of the spaces. The key novelty of the proposed approach is to utilise the relationship between heat kernels and the transition density of Brownian motion on manifolds for constructing and approximating valid and computationally feasible covariance kernels. This enables in-GPs to be practically applied in great generality, while existing approaches for smoothing on constrained domains are limited to simple special cases. The broad utilities of the in-GP approach is illustrated through simulation studies and data examples.

[1]  Wei Cai,et al.  Computation of Local Time of Reflecting Brownian Motion and Probabilistic Representation of the Neumann Problem , 2015, 1502.01319.

[2]  Carl E. Rasmussen,et al.  A Unifying View of Sparse Approximate Gaussian Process Regression , 2005, J. Mach. Learn. Res..

[3]  Carl E. Rasmussen,et al.  Gaussian processes for machine learning , 2005, Adaptive computation and machine learning.

[4]  Bernard Lapeyre,et al.  Introduction to Stochastic Calculus Applied to Finance , 2007 .

[5]  J. Weston,et al.  Approximation Methods for Gaussian Process Regression , 2007 .

[6]  P. Guttorp,et al.  Nonparametric Estimation of Nonstationary Spatial Covariance Structure , 1992 .

[7]  Van Der Vaart,et al.  Adaptive Bayesian estimation using a Gaussian random field with inverse Gamma bandwidth , 2009, 0908.3556.

[8]  Zoubin Ghahramani,et al.  Sparse Gaussian Processes using Pseudo-inputs , 2005, NIPS.

[9]  Simon N. Wood,et al.  Soap film smoothing , 2008 .

[10]  Bruno Pelletier Kernel density estimation on Riemannian manifolds , 2005 .

[11]  John Sylvester,et al.  The heat equation and reflected Brownian motion in time-dependent domains , 2004 .

[12]  M. Katzfuss,et al.  A General Framework for Vecchia Approximations of Gaussian Processes , 2017, 1708.06302.

[13]  Kaleem Siddiqi,et al.  Stochastic Heat Kernel Estimation on Sampled Manifolds , 2017, Comput. Graph. Forum.

[14]  David Dunson,et al.  Extrinsic Gaussian Processes for Regression and Classification on Manifolds , 2017, Bayesian Analysis.

[15]  Montserrat Fuentes,et al.  Isotropic covariance functions on spheres: Some properties and modeling considerations , 2016, J. Multivar. Anal..

[16]  P. Lions,et al.  Stochastic differential equations with reflecting boundary conditions , 1984 .

[17]  Tim Ramsay,et al.  Spline smoothing over difficult regions , 2002 .

[18]  Anton Schwaighofer,et al.  Transductive and Inductive Methods for Approximate Gaussian Process Regression , 2002, NIPS.

[19]  David B. Dunson,et al.  Nonparametric Bayes regression and classification through mixtures of product kernels , 2010 .

[20]  G. Matheron The intrinsic random functions and their applications , 1973, Advances in Applied Probability.

[21]  Gerard Kerkyacharian,et al.  Thomas Bayes’ walk on manifolds , 2012, 1206.0459.

[22]  Michalis K. Titsias,et al.  Variational Learning of Inducing Variables in Sparse Gaussian Processes , 2009, AISTATS.

[23]  P. Kloeden,et al.  Higher-order implicit strong numerical schemes for stochastic differential equations , 1992 .

[24]  市原 完治 Brownian motion on a Riemannian manifold , 1981 .

[25]  James O. Ramsay,et al.  Spatial spline regression models , 2013 .

[26]  M. N. Ndumu Brownian Motion and Riemannian Geometry in the Neighbourhood of a Submanifold , 2011 .