pith. sign in

arxiv: 2605.21307 · v1 · pith:5ICBSBBAnew · submitted 2026-05-20 · 📊 stat.ME

The Bayesian Gaussian Process Latent Variable Model for Spatio-Temporal Stream Networks

Marno Basson , Tobias M. Louw , Theresa R. Smith This is my paper

Pith reviewed 2026-05-21 03:40 UTC · model grok-4.3

classification 📊 stat.ME
keywords Gaussian processeslatent variable modelsspatio-temporal modelingstream networksvariational inferencecensored datacovariance functions
0
0 comments X

The pith

A variational inference framework trains multi-output Gaussian process latent variable models for spatio-temporal stream networks using stream distances on censored data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a variational inference-based framework to train a multi-output Gaussian process latent variable model specifically for tails-up spatio-temporal stream networks. It replaces Euclidean distance with stream distance and builds separable covariance functions through auto/cross-correlation for spatial links plus process convolution for temporal links. Training maximizes a secondary variational lower bound on the log marginal likelihood to handle censored observations that include missing values. A sympathetic reader would care because this supplies a practical route to modeling incomplete environmental measurements along river or stream systems where flow paths matter more than straight-line separation.

Core claim

The paper claims that a new family of tails-up spatio-temporal stream network models arises from combining the sparse Gaussian process inducing variable framework, the Bayesian Gaussian process latent variable model, and local variational methods. These models use stream distance rather than Euclidean distance and capture spatial and temporal dependencies via auto/cross-correlation and process convolution, respectively, which yields valid separable spatio-temporal stream network-based covariance functions. Training on a censored observational data set subject to missing values proceeds by maximising a secondary variational lower bound on the model log marginal likelihood using gradient-based

What carries the argument

The tails-up spatio-temporal stream network covariance function built from stream distance, auto/cross-correlation, and process convolution inside the Bayesian Gaussian process latent variable model with sparse inducing variables and local variational inference.

If this is right

  • The resulting models support multi-output prediction on stream networks while respecting flow topology.
  • Simulation studies demonstrate competitive performance against standard benchmarks on several metrics.
  • The framework directly accommodates missing values without requiring separate imputation steps.
  • Valid covariance functions become available for any separable spatio-temporal stream network setting.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same variational construction could extend to other linear network structures such as road or utility grids.
  • Replacing the inducing point approximation with more scalable sparse methods might allow application to very large monitoring networks.
  • Downstream tasks such as anomaly detection in water quality time series could use the latent variables directly.

Load-bearing premise

That stream distance combined with auto/cross-correlation and process convolution produces valid separable spatio-temporal covariance functions that remain trainable by maximising a secondary variational lower bound on censored data.

What would settle it

Apply the trained model to a held-out real stream network dataset with known complete observations and check whether its predictive accuracy on missing or censored points exceeds that of an otherwise identical model using Euclidean distance.

read the original abstract

A variational inference-based framework for training a multi-output Gaussian process latent variable model, specifically tailored to the tails-up spatio-temporal stream network, is developed. Training, given a censored observational data set subject to missing values, proceeds by maximising a secondary variational lower bound on the model log marginal likelihood using gradient-based optimisation. Consequently, the theoretical development for a new family of tails-up spatio-temporal stream network models is introduced which rely on the sparse Gaussian process inducing variable framework, the Bayesian Gaussian process latent variable model, and local variational methods. These spatio-temporal models use stream distance instead of Euclidean distance and capture spatial and temporal dependencies using auto/cross-correlation and process convolution, respectively, which allows for the development of valid separable spatio-temporal stream network-based covariance functions. Results from the simulation-based case studies indicate that the proposed framework performs well when considering benchmark comparisons and several performance metrics.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The paper develops a variational inference framework for a multi-output Bayesian Gaussian Process Latent Variable Model (GPLVM) tailored to tails-up spatio-temporal stream networks. It constructs a new family of models by replacing Euclidean distance with stream distance in spatial auto- and cross-correlation functions, combined with process convolution for temporal components to yield separable spatio-temporal covariances. Training maximizes a secondary variational lower bound (ELBO) on the log marginal likelihood for censored data with missing values, using sparse GP inducing variables and local variational methods. Performance is assessed via simulation-based case studies against benchmarks using multiple metrics.

Significance. If the stream-distance-based kernels are provably valid and the variational scheme is reliable, the framework would offer a principled extension of GPLVMs to directed acyclic network domains with censored observations, relevant for applications in hydrology and environmental statistics. The simulation results provide initial evidence of practical performance, but significance hinges on verification of the core covariance construction.

major comments (1)
  1. [Theoretical development section] Theoretical development section: The manuscript asserts that stream distance combined with auto/cross-correlation (spatial) and process convolution (temporal) yields valid separable spatio-temporal stream network covariance functions, yet provides no explicit proof that the Gram matrix remains positive semi-definite for arbitrary stream network topologies, nor any numerical check on a non-trivial directed acyclic graph. This property is load-bearing for the central claim, as an invalid kernel would render the Gaussian process model and the subsequent maximization of the secondary ELBO on censored data ill-posed.
minor comments (2)
  1. [Abstract] Abstract and introduction: More detail is needed on the specific form of the local variational approximation, handling of censored observations, and any potential weaknesses or convergence issues in the gradient-based optimization of the secondary ELBO.
  2. [Simulation-based case studies] Simulation studies: Clarify the exact benchmark methods, network topologies used for validation, and whether any checks for positive-definiteness of the constructed kernels were performed during the case studies.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their careful reading of the manuscript and for identifying a key point regarding the theoretical validity of the proposed covariance functions. We address this concern directly below and will revise the manuscript to incorporate additional supporting material.

read point-by-point responses

  1. Referee: Theoretical development section: The manuscript asserts that stream distance combined with auto/cross-correlation (spatial) and process convolution (temporal) yields valid separable spatio-temporal stream network covariance functions, yet provides no explicit proof that the Gram matrix remains positive semi-definite for arbitrary stream network topologies, nor any numerical check on a non-trivial directed acyclic graph. This property is load-bearing for the central claim, as an invalid kernel would render the Gaussian process model and the subsequent maximization of the secondary ELBO on censored data ill-posed.

    Authors: We thank the referee for highlighting this important requirement. The manuscript develops the spatio-temporal covariance by substituting stream distance into established positive definite auto- and cross-correlation functions for the spatial component and employing process convolution for the temporal component, yielding a separable form. We acknowledge, however, that the current version does not contain an explicit proof of positive semi-definiteness of the resulting Gram matrix for arbitrary directed acyclic stream network topologies, nor numerical verification on non-trivial graphs. In the revised manuscript we will add a formal argument in the Theoretical development section establishing that the construction preserves positive definiteness when stream distance is used as the metric (leveraging the fact that the component kernels are positive definite on the induced metric space), together with numerical checks confirming that Gram matrices remain positive semi-definite on several non-trivial DAG examples. These additions will directly support the central modeling claim. revision: yes

Circularity Check

0 steps flagged

No significant circularity; derivation builds on established external GP components

full rationale

The paper presents a variational framework for a multi-output Bayesian GPLVM tailored to tails-up spatio-temporal stream networks. It explicitly relies on the sparse GP inducing variable framework, the Bayesian GPLVM, and local variational methods as foundational building blocks, then combines them with stream distance, auto/cross-correlation, and process convolution to form separable covariance functions. No equation or claim in the abstract reduces a central result to a fitted parameter or self-defined quantity by construction, nor does any load-bearing step invoke a uniqueness theorem or ansatz justified solely by the authors' own prior unverified work. The development is therefore self-contained against external benchmarks in Gaussian process literature.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Based solely on the abstract, the model rests on standard domain assumptions in Gaussian process modeling for networks; no free parameters or invented entities are explicitly detailed.

axioms (2)
  • domain assumption Stream distance instead of Euclidean distance yields valid covariance functions for network data
    Invoked when developing the spatio-temporal covariance functions
  • domain assumption Process convolution captures temporal dependencies in a separable manner
    Used to construct the spatio-temporal models

pith-pipeline@v0.9.0 · 5681 in / 1225 out tokens · 37249 ms · 2026-05-21T03:40:22.513386+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

109 extracted references · 109 canonical work pages · 1 internal anchor

  1. [1]

    Journal of the American Statistical Association , number =

    Ver Hoef, Jay M and Peterson, Erin E , doi =. Journal of the American Statistical Association , number =. 2010 , bdsk-url-1 =

    work page 2010

  2. [2]

    Computational Statistics & Data Analysis , keywords =

    Santos-Fernandez, Edgar and. Computational Statistics & Data Analysis , keywords =. 2022 , bdsk-url-1 =. doi:https://doi.org/10.1016/j.csda.2022.107446 , issn =

    work page doi:10.1016/j.csda.2022.107446 2022

  3. [3]

    Environmental and Ecological Statistics , number =

    Ver Hoef, Jay M and Peterson, Erin and Theobald, David , doi =. Environmental and Ecological Statistics , number =. 2006 , bdsk-url-1 =

    work page 2006

  4. [4]

    Statistics and Computing , number =

    Basson, Marno and Louw, Tobias M and Smith, Theresa R , doi =. Statistics and Computing , number =. 2023 , bdsk-url-1 =

    work page 2023

  5. [5]

    Mihelcic, James R and Zimmerman, Julie Beth , edition =

    work page

  6. [6]

    and Skopec, Mary and Meyer, Michael T

    Kolpin, Dana W. and Skopec, Mary and Meyer, Michael T. and Furlong, Edward T. and Zaugg, Steven D. , doi =. Science of the Total Environment , keywords =. 2004 , bdsk-url-1 =

    work page 2004

  7. [7]

    and Furlong, Edward T

    Kolpin, Dana W. and Furlong, Edward T. and Meyer, Michael T. and Thurman, E. Michael and Zaugg, Steven D. and Barber, Larry B. and Buxton, Herbert T. , doi =. Environmental Science and Technology , title =. 2002 , bdsk-url-1 =

    work page 2002

  8. [8]

    Archer, Edward and Petrie, Bruce and Kasprzyk-Hordern, Barbara and Wolfaardt, Gideon M. , doi =. Chemosphere , keywords =. 2017 , bdsk-url-1 =

    work page 2017

  9. [9]

    and van Wyk, Johannes H

    Archer, Edward and Wolfaardt, Gideon M. and van Wyk, Johannes H. , doi =. Water SA , keywords =. 2017 , bdsk-url-1 =

    work page 2017

  10. [10]

    and Furlong, Edward T

    Glassmeyer, Susan T. and Furlong, Edward T. and Kolpin, Dana W. and Cahill, Jeffery D. and Zaugg, Steven D. and Werner, Stephen L. and Meyer, Michael T. and Kryak, David D. , doi =. Environmental Science and Technology , number =. 2005 , bdsk-url-1 =

    work page 2005

  11. [11]

    2019 , bdsk-url-1 =

    Lorenzo, Maria and Pic. 2019 , bdsk-url-1 =. doi:https://doi.org/10.1016/j.coesh.2019.05.007 , journal =

    work page doi:10.1016/j.coesh.2019.05.007 2019

  12. [12]

    and Pleus, Richard C

    Bruce, Gretchen M. and Pleus, Richard C. and Snyder, Shane A. , doi =. Toxicological Relevance of Pharmaceuticals in Drinking Water , volume =. Environmental Science & Technology , number =. 2010 , bdsk-url-1 =

    work page 2010

  13. [13]

    and Mulukutla, G.K

    Wollheim, W.M. and Mulukutla, G.K. and Cook, C. and Carey, R.O. , doi =. Aquatic Nitrate Retention at River Network Scales Across Flow Conditions Determined Using Nested In Situ Sensors , volume =. Water Resources Research , number =. 2017 , bdsk-url-1 =

    work page 2017

  14. [14]

    and Stets, Edward G

    Stackpoole, Sarah M. and Stets, Edward G. and Clow, David W. and Burns, Douglas A. and Aiken, George R. and Aulenbach, Brent T. and Creed, Irena F. and Hirsch, Robert M. and Laudon, Hjalmar and Pellerin, Brian A. and Striegl, Robert G. , doi =. Spatial and temporal patterns of dissolved organic matter quantity and quality in the Mississippi River Basin, 1...

    work page 1997

  15. [15]

    and Ver Hoef, Jay M

    Peterson, Erin E. and Ver Hoef, Jay M. , doi =. A mixed-model moving-average approach to geostatistical modeling in stream networks , volume =. Ecology , keywords =. 2010 , bdsk-url-1 =

    work page 2010

  16. [16]

    and Theobald, David M

    Peterson, Erin E. and Theobald, David M. and Ver Hoef, Jay M. , doi =. Geostatistical modelling on stream networks: developing valid covariance matrices based on hydrologic distance and stream flow , volume =. Freshwater Biology , keywords =. 2007 , bdsk-url-1 =

    work page 2007

  17. [17]

    Groot, Perry and Lucas, Peter , booktitle =

    work page

  18. [18]

    Econometrica , number =

    Tobin, James , doi =. Econometrica , number =. 1958 , bdsk-url-1 =

    work page 1958

  19. [19]

    2007 , bdsk-url-1 =

    Ertin, Emre , booktitle =. 2007 , bdsk-url-1 =. doi:10.1109/SSP.2007.4301342 , pages =

    work page doi:10.1109/ssp.2007.4301342 2007

  20. [20]

    Generalized multi-output Gaussian process censored regression , volume =

    Daniele Gammelli and Kasper Pryds Rolsted and Dario Pacino and Filipe Rodrigues , doi =. Generalized multi-output Gaussian process censored regression , volume =. Pattern Recognition , keywords =. 2022 , bdsk-url-1 =

    work page 2022

  21. [21]

    C , keywords =

    Cressie, Noel A. C , keywords =

    work page

  22. [22]

    and Patterson, David A

    Hague, Merran J. and Patterson, David A. , doi =. Evaluation of Statistical River Temperature Forecast Models for Fisheries Management , volume =. North American Journal of Fisheries Management , number =. 2014 , bdsk-url-1 =

    work page 2014

  23. [23]

    Distribution properties of a measurement series of river water temperature at different time resolution levels (based on the example of the Lowland River Note

    Graf, Renata , doi =. Distribution properties of a measurement series of river water temperature at different time resolution levels (based on the example of the Lowland River Note. Water , number =. 2018 , bdsk-url-1 =

    work page 2018

  24. [24]

    Spatial prediction on a river network , volume =

    Cressie, Noel and Frey, Jesse and Harch, Bronwyn and Smith, Mick , doi =. Spatial prediction on a river network , volume =. Journal of Agricultural, Biological, and Environmental Statistics , number =. 2006 , bdsk-url-1 =

    work page 2006

  25. [25]

    and D'Amico, Ellen and Smith, Elizabeth M

    McManus, Michael G. and D'Amico, Ellen and Smith, Elizabeth M. and Polinsky, Robyn and Ackerman, Jerry and Tyler, Kip , doi =. Variation in stream network relationships and geospatial predictions of watershed conductivity , volume =. Freshwater Science , number =. 2020 , bdsk-url-1 =

    work page 2020

  26. [26]

    and Carter, Gail P

    Money, Eric S. and Carter, Gail P. and Serre, Marc L. , doi =. Modern Space/Time Geostatistics Using River Distances: Data Integration of Turbidity and E. coli Measurements to Assess Fecal Contamination Along the Raritan River in New Jersey , volume =. Environmental Science & Technology , number =. 2009 , bdsk-url-1 =

    work page 2009

  27. [27]

    and Carter, Gail P

    Money, Eric S. and Carter, Gail P. and Serre, Marc L. , doi =. Using river distances in the space/time estimation of dissolved oxygen along two impaired river networks in New Jersey , volume =. Water Research , keywords =. 2009 , bdsk-url-1 =

    work page 2009

  28. [28]

    and Marian Scott, E

    O'Donnell, David and Rushworth, Alastair and Bowman, Adrian W. and Marian Scott, E. and Hallard, Mark , doi =. Journal of the Royal Statistical Society Series C: Applied Statistics , month =. 2014 , bdsk-url-1 =

    work page 2014

  29. [29]

    Jackson and Robert J

    Faye L. Jackson and Robert J. Fryer and David M. Hannah and Colin P. Millar and Iain A. Malcolm , doi =. A spatio-temporal statistical model of maximum daily river temperatures to inform the management of Scotland's Atlantic salmon rivers under climate change , volume =. Science of The Total Environment , keywords =. 2018 , bdsk-url-1 =

    work page 2018

  30. [30]

    and Messier, Kyle P

    Holcomb, David A. and Messier, Kyle P. and Serre, Marc L. and Rowny, Jakob G. and Stewart, Jill R. , doi =. Geostatistical Prediction of Microbial Water Quality Throughout a Stream Network Using Meteorology, Land Cover, and Spatiotemporal Autocorrelation , volume =. Environmental Science & Technology , number =. 2018 , bdsk-url-1 =

    work page 2018

  31. [31]

    and Peterson, Erin E

    Isaak, Daniel J. and Peterson, Erin E. and Ver Hoef, Jay M. and Wenger, Seth J. and Falke, Jeffrey A. and Torgersen, Christian E. and Sowder, Colin and Steel, E. Ashley and Fortin, Marie-Josee and Jordan, Chris E. and Ruesch, Aaron S. and Som, Nicholas and Monestiez, Pascal , doi =. Applications of spatial statistical network models to stream data , volum...

    work page 2014

  32. [32]

    Using spatial-stream-network models and long-term data to understand and predict dynamics of faecal contamination in a mixed land-use catchment , volume =

    Aaron James Neill and Doerthe Tetzlaff and Norval James Colin Strachan and Rupert Lloyd Hough and Lisa Marie Avery and Helen Watson and Chris Soulsby , doi =. Using spatial-stream-network models and long-term data to understand and predict dynamics of faecal contamination in a mixed land-use catchment , volume =. Science of The Total Environment , keyword...

    work page 2018

  33. [33]

    and Lawrence, Neil D

    Titsias, Michalis K. and Lawrence, Neil D. , booktitle =. Bayesian Gaussian Process Latent Variable Model , volume =

    work page

  34. [34]

    Chantal. Mod. Comptes Rendus Geoscience , keywords =. 2006 , bdsk-url-1 =. doi:https://doi.org/10.1016/j.crte.2006.02.002 , issn =

    work page doi:10.1016/j.crte.2006.02.002 2006

  35. [35]

    arXiv , author =:2009.14745 , title =

    work page arXiv 2009

  36. [36]

    Daily River Water Temperature Prediction: A Comparison between Neural Network and Stochastic Techniques , volume =

    Graf, Renata and Aghelpour, Pouya , doi =. Daily River Water Temperature Prediction: A Comparison between Neural Network and Stochastic Techniques , volume =. Atmosphere , number =. 2021 , bdsk-url-1 =

    work page 2021

  37. [37]

    Bishop, Christopher M , keywords =

    work page

  38. [38]

    Rasmussen, Carl Edward and Williams, Christopher K.I , keywords =

    work page

  39. [39]

    Garreta, Vincent and Monestiez, Pascal and Ver Hoef, Jay M. , doi =. Spatial modelling and prediction on river networks: up model, down model or hybrid? , volume =. Environmetrics , keywords =. 2010 , bdsk-url-1 =

    work page 2010

  40. [40]

    Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics , pages =

    Efficient Multioutput Gaussian Processes through Variational Inducing Kernels , volume =. Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics , pages =

    work page

  41. [41]

    and Titsias, Michalis K

    Damianou, Andreas C. and Titsias, Michalis K. and Lawrence, Neil D. , journal =

    work page

  42. [42]

    Titsias , booktitle =

    Michalis K. Titsias , booktitle =. Variational

    work page

  43. [43]

    Titsias , institution =

    Michalis K. Titsias , institution =. 2008 , bdsk-url-1 =

    work page 2008

  44. [44]

    An introduction to variational methods for graphical models , volume =

    Jordan, Michael I and Ghahramani, Zoubin and Jaakkola, Tommi S and Saul, Lawrence K , doi =. An introduction to variational methods for graphical models , volume =. Machine learning , number =. 1999 , bdsk-url-1 =

    work page 1999

  45. [45]

    Approximations for

    Hannes Nickisch and Carl Edward Rasmussen , journal =. Approximations for

    work page

  46. [46]

    Titsias, Michalis K. and L\'. Variational Inference for Mahalanobis Distance Metrics in Gaussian Process Regression , year =. Proceedings of the 26th International Conference on Neural Information Processing Systems - Volume 1 , pages =

    work page

  47. [47]

    and Lawrence, Neil D

    Damianou, Andreas C. and Lawrence, Neil D. , booktitle =. Deep

    work page

  48. [48]

    Variational Heteroscedastic Gaussian Process Regression , year =

    L\'. Variational Heteroscedastic Gaussian Process Regression , year =. Proceedings of the 28th International Conference on International Conference on Machine Learning , isbn =

    work page

  49. [49]

    and Lawrence, Neil D

    Damianou, Andreas C and Titsias, Michalis K. and Lawrence, Neil D. , journal =. Variational Gaussian process dynamical systems , volume =

    work page

  50. [50]

    Variational Dependent Multi-output Gaussian Process Dynamical Systems , url =

    Jing Zhao and Shiliang Sun , journal =. Variational Dependent Multi-output Gaussian Process Dynamical Systems , url =. 2016 , bdsk-url-1 =

    work page 2016

  51. [51]

    Gaussian Process Training with Input Noise , year =

    McHutchon, Andrew and Rasmussen, Carl Edward , booktitle =. Gaussian Process Training with Input Noise , year =

    work page

  52. [52]

    and Rasmussen, C

    Girard, A. and Rasmussen, C. E. and Candela, J. Qui. Gaussian Process Priors with Uncertain Inputs - Application to Multiple-Step Ahead Time Series Forecasting , year =. Advances in Neural Information Processing Systems 15 , edition =

    work page

  53. [53]

    and Kucukelbir, Alp and McAuliffe, Jon D

    Blei, David M. and Kucukelbir, Alp and McAuliffe, Jon D. , doi =. Journal of the American Statistical Association , keywords =. 2017 , bdsk-url-1 =

    work page 2017

  54. [54]

    Wainwright and Michael I

    Martin J. Wainwright and Michael I. Jordan , doi =. Graphical Models, Exponential Families, and Variational Inference , volume =. Foundations and Trends. 2008 , bdsk-url-1 =

    work page 2008

  55. [55]

    MacKay, David J C , keywords =

    work page

  56. [56]

    Statistical Field Theory , year =

    Parisi, Giorgio , pages =. Statistical Field Theory , year =

    work page

  57. [57]

    Damianou , publisher =

    Andreas C. Damianou , publisher =. Deep Gaussian Processes and Variational Propagation of Uncertainty , year =

    work page

  58. [58]

    2002 , bdsk-url-1 =

    Csat. 2002 , bdsk-url-1 =. doi:10.1162/089976602317250933 , journal =

    work page doi:10.1162/089976602317250933 2002

  59. [59]

    Seeger, Matthias and Williams, Christopher K I and Lawrence, Neil D , booktitle =

    work page

  60. [60]

    Sparse Gaussian processes using pseudo-inputs , volume =

    Snelson, Edward and Ghahramani, Zoubin , journal =. Sparse Gaussian processes using pseudo-inputs , volume =

    work page

  61. [61]

    Joaquin Qui. A. Journal of Machine Learning Research , number =

    work page

  62. [62]

    Probabilistic Non-linear Principal Component Analysis with Gaussian Process Latent Variable Models , volume =

    Neil Lawrence , journal =. Probabilistic Non-linear Principal Component Analysis with Gaussian Process Latent Variable Models , volume =

    work page

  63. [63]

    Gaussian Process Latent Variable Models for Visualisation of High Dimensional Data , volume =

    Lawrence, Neil , booktitle =. Gaussian Process Latent Variable Models for Visualisation of High Dimensional Data , volume =

    work page

  64. [64]

    , booktitle =

    Lawrence, Neil D. , booktitle =. Learning for Larger Datasets with the Gaussian Process Latent Variable Model , volume =

    work page

  65. [65]

    A review on Gaussian Process Latent Variable Models , volume =

    Ping Li and Songcan Chen , doi =. A review on Gaussian Process Latent Variable Models , volume =. CAAI Transactions on Intelligence Technology , number =. 2016 , bdsk-url-1 =

    work page 2016

  66. [66]

    2014 , bdsk-url-1 =

    Bui, Thang and Turner, Richard , title =. 2014 , bdsk-url-1 =

    work page 2014

  67. [67]

    Journal of the American Statistical Association , number =

    Berkson, Joseph , doi =. Journal of the American Statistical Association , number =. 1950 , bdsk-url-1 =

    work page 1950

  68. [68]

    The Annals of Statistics , number =

    Schennach, Susanne M , doi =. The Annals of Statistics , number =. 2013 , bdsk-url-1 =

    work page 2013

  69. [69]

    2009 , bdsk-url-1 =

    Opper, Manfred and Archambeau, C. 2009 , bdsk-url-1 =. doi:10.1162/neco.2008.08-07-592 , journal =

    work page doi:10.1162/neco.2008.08-07-592 2009

  70. [70]

    Variational algorithms for approximate Bayesian inference , year =

    Beal, Matthew James , publisher =. Variational algorithms for approximate Bayesian inference , year =

    work page

  71. [71]

    Variational message passing and its applications , year =

    Winn, John Michael , publisher =. Variational message passing and its applications , year =

    work page

  72. [72]

    Variational message passing

    Winn, John and Bishop, Christopher M and Jaakkola, Tommi , journal =. Variational message passing. , volume =

    work page

  73. [73]

    Variational Inference with Normalizing Flows , volume =

    Rezende, Danilo and Mohamed, Shakir , booktitle =. Variational Inference with Normalizing Flows , volume =

    work page

  74. [74]

    Normalizing flows for probabilistic modeling and inference , volume =

    Papamakarios, George and Nalisnick, Eric and Rezende, Danilo Jimenez and Mohamed, Shakir and Lakshminarayanan, Balaji , journal =. Normalizing flows for probabilistic modeling and inference , volume =

    work page

  75. [75]

    Gaussian Process Latent Variable Flows for Massively Missing Data , year =

    Vidhi Lalchand and Aditya Ravuri and Neil D Lawrence , booktitle =. Gaussian Process Latent Variable Flows for Massively Missing Data , year =

    work page

  76. [76]

    Copula variational inference , volume =

    Tran, Dustin and Blei, David and Airoldi, Edo M , booktitle =. Copula variational inference , volume =

    work page

  77. [77]

    Fast copula variational inference , volume =

    Jinjin Chi and Jihong Ouyang and Ang Zhang and Xinhua Wang and Ximing Li , doi =. Fast copula variational inference , volume =. Journal of Experimental & Theoretical Artificial Intelligence , number =. 2022 , bdsk-url-1 =

    work page 2022

  78. [78]

    Survival Regression with Accelerated Failure Time Model in XGBoost

    Michael Stanley Smith and Rub. Implicit Copula Variational Inference , volume =. 2022 , bdsk-url-1 =. doi:10.1080/10618600.2022.2119987 , journal =

    work page doi:10.1080/10618600.2022.2119987 2022

  79. [79]

    Variational inference with Gaussian mixture model and householder flow , volume =

    GuoJun Liu and Yang Liu and MaoZu Guo and Peng Li and MingYu Li , doi =. Variational inference with Gaussian mixture model and householder flow , volume =. Neural Networks , keywords =. 2019 , bdsk-url-1 =

    work page 2019

  80. [80]

    Logan, J D , edition =

    work page

Showing first 80 references.