pith. sign in

arxiv: 2606.23326 · v1 · pith:3252FPQ4new · submitted 2026-06-22 · 📊 stat.ME · stat.AP

Online forecast reconciliation using linear models

Tobias R{\o}nlev-Knudsen , Henrik Madsen , Jan Kloppenborg M{\o}ller This is my paper

Pith reviewed 2026-06-26 07:35 UTC · model grok-4.3

classification 📊 stat.ME stat.AP
keywords forecast reconciliationhierarchical forecastingonline learningridge regressionmatrix normal distributiontemporal hierarchydistrict heating
0
0 comments X

The pith

A multivariate linear model with ridge regression enables online hierarchical forecast reconciliation.

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

The paper formalizes forecast hierarchies as graphs and casts reconciliation as estimation in a multivariate linear model whose residuals follow a matrix normal distribution. Parameters are found by ridge regression, which yields both point estimates and uncertainty measures, then updated recursively without reprocessing all past data. The resulting scheme produces reconciled forecasts that adapt as new observations arrive. A reader would care because many practical forecasting problems, such as energy loads, operate in continuous time and cannot wait for batch recomputation of reconciled values.

Core claim

Hierarchies can be reconciled online by embedding the reconciliation step inside a multivariate linear model whose residuals are matrix-normal; ridge regression supplies the parameter estimates and a recursive least-squares update rule then propagates those estimates forward in time, giving both reconciled point forecasts and their uncertainty at every step.

What carries the argument

Multivariate linear model with matrix-normal residuals, estimated by ridge regression and updated by a recursive least-squares scheme.

If this is right

  • Reconciled forecasts become available after each new observation without recomputing from scratch.
  • Uncertainty intervals for both parameters and reconciled forecasts can be tracked continuously.
  • Shrinkage induced by the ridge penalty improves reconciliation stability when the hierarchy is deep.
  • The same recursive machinery applies to any hierarchy that can be written as a graph.

Where Pith is reading between the lines

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

  • The recursive formulation may cut computation time for very large or frequently updated hierarchies.
  • The framework could be combined with other online learners that supply the initial forecasts.
  • Temporal hierarchies in energy systems may benefit most because load patterns evolve smoothly.

Load-bearing premise

The residuals across all levels of the hierarchy are adequately described by a single matrix normal distribution.

What would settle it

Run the recursive updates on the district-heating temporal hierarchy and compare the online reconciled forecasts against the batch ridge-regression solution; large systematic divergence would falsify the online scheme.

Figures

Figures reproduced from arXiv: 2606.23326 by Henrik Madsen, Jan Kloppenborg M{\o}ller, Tobias R{\o}nlev-Knudsen.

Figure 1
Figure 1. Figure 1: A hierarchy and its summation matrix. Nodes represent states with values con [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic of the implementation of forecast reconciliation in the [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Heat load data provided by Kredsløb after removing outliers and resampling to [PITH_FULL_IMAGE:figures/full_fig_p014_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Temporal hierarchy used for load forecasting. The bottom level considers horizons [PITH_FULL_IMAGE:figures/full_fig_p014_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Local temperature measurements. east region and applied to all three regions. A linear model based on section 3.2 was used to estimate parameters and generate forecasts. For further details on the base forecasts, refer to Appendix D. Case Update frequency θ0 Q λmem Main Hourly 0 0.001 0.995 1 Daily 0 0.001 0.9952 1 3 100 [PITH_FULL_IMAGE:figures/full_fig_p015_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Sample of reconciled forecasts for the east region including prediction bands as [PITH_FULL_IMAGE:figures/full_fig_p016_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Relative reduction in RMSE achieved through forecast reconciliation (main case). [PITH_FULL_IMAGE:figures/full_fig_p017_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: RRMSE of reconciled forecasts by aggregation level (main case). [PITH_FULL_IMAGE:figures/full_fig_p017_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: RRMSE scores for cases 1-3 for bottom level reconciled forecasts compared to base [PITH_FULL_IMAGE:figures/full_fig_p019_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: MSE and averaged forecasts of prediction error variance estimates for the east [PITH_FULL_IMAGE:figures/full_fig_p019_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Sample of forecasts for the east region and all aggregation levels. Base forecasts [PITH_FULL_IMAGE:figures/full_fig_p033_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: RMSE score for base forecasts of each region. [PITH_FULL_IMAGE:figures/full_fig_p034_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Fraction of variance explained by reconciled forecasts ( [PITH_FULL_IMAGE:figures/full_fig_p034_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: RRMSE for increasing regularization for each region, aggregation level, and [PITH_FULL_IMAGE:figures/full_fig_p035_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: MSE and averaged forecasts of prediction error variance estimates for the south [PITH_FULL_IMAGE:figures/full_fig_p035_15.png] view at source ↗
read the original abstract

We present a framework for online and adaptive forecasting and hierarchical reconciliation using linear regression models. We begin by formalizing hierarchies using graphs, and motivated by their structure, formulate a multivariate linear model using the matrix normal distribution to characterize residuals. Parameter estimation is posed as a ridge regression problem and applied to hierarchical forecast reconciliation. The connections between ridge regression, Bayesian estimation and shrinkage for hierarchical reconciliation are discussed, and results for uncertainty quantification in parameters and forecasts are provided. Based on the ridge regression formulation, a recursive inference scheme inspired by recursive least squares is described. The algorithm is implemented in the PyOnlineForecast package. Finally, the proposed methodology is demonstrated on a case study for district heating load forecasting using a temporal hierarchy. Our results provide a reference for implementation of forecast reconciliation via multivariate linear models in an online setting. The case study furthermore highlights practical considerations of using temporal hierarchies in an online setting and demonstrates the usefulness of the proposed framework and implementation, both for district heating load forecasting and more generally for online hierarchical forecasting.

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

0 major / 3 minor

Summary. The paper presents a framework for online and adaptive forecasting and hierarchical reconciliation using linear regression models. Hierarchies are formalized via graphs, a multivariate linear model with matrix-normal residuals is introduced, estimation is formulated as ridge regression (with links to Bayesian shrinkage), a recursive least-squares-style update is derived, uncertainty quantification is provided, the method is implemented in PyOnlineForecast, and it is demonstrated on a temporal-hierarchy case study for district heating load forecasting.

Significance. If the derivations and empirical results hold, the work supplies a coherent, implementable reference for performing hierarchical reconciliation inside an online linear-model setting. Explicit connections among ridge regression, Bayesian estimation, and shrinkage, together with the recursive update and open-source package, constitute concrete strengths for reproducibility and practical adoption in streaming hierarchical forecasting tasks.

minor comments (3)
  1. [Abstract] The abstract states that 'results for uncertainty quantification in parameters and forecasts are provided,' yet the precise form (e.g., analytic variance expressions, credible intervals, or bootstrap procedures) is not indicated; a one-sentence clarification would help readers locate the relevant derivation.
  2. The transition from the graph-based hierarchy definition to the matrix-normal multivariate model would benefit from an explicit low-dimensional example showing how the graph adjacency determines the residual covariance structure.
  3. The case-study section should report the numerical values of the ridge regularization parameter(s) used and any sensitivity checks performed, as this parameter is listed among the free parameters.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for their positive assessment of the manuscript, including the summary of the framework, the noted strengths in connections to ridge regression and Bayesian estimation, the recursive update, and the open-source implementation. We appreciate the recommendation for minor revision.

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The derivation formalizes hierarchies as graphs, adopts the matrix-normal distribution as a modeling assumption for residuals, casts estimation as ridge regression, and derives a recursive update from standard recursive least-squares. All components are drawn from established statistical literature and applied to reconciliation; none of the load-bearing steps (graph formalization, ridge objective, recursive scheme, or uncertainty quantification) reduce by definition or self-citation to quantities defined by the same model. The framework remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The framework rests on standard linear-model assumptions plus the choice of ridge regularization; no new entities are postulated.

free parameters (1)
  • ridge regularization parameter
    Introduced to stabilize parameter estimation in the ridge regression step; its specific value is not given in the abstract.
axioms (2)
  • domain assumption Hierarchies can be formalized using graphs
    Invoked to motivate the structure of the multivariate linear model.
  • domain assumption Residuals follow a matrix normal distribution
    Used to characterize the joint distribution of errors across the hierarchy.

pith-pipeline@v0.9.1-grok · 5709 in / 1161 out tokens · 26905 ms · 2026-06-26T07:35:54.614767+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

85 extracted references · 61 canonical work pages · 4 internal anchors

  1. [1]

    Stochastic Processes and their Applications , author =

    Reverse-Time Diffusion Equation Models , author =. doi:10.1016/0304-4149(82)90051-5 , url =

    work page doi:10.1016/0304-4149(82)90051-5

  2. [2]

    doi:10.1016/j.ejor.2017.02.046 , url =

    Forecasting with Temporal Hierarchies , author =. doi:10.1016/j.ejor.2017.02.046 , url =

    work page doi:10.1016/j.ejor.2017.02.046 2017

  3. [3]

    and Hyndman, Rob J

    Athanasopoulos, George and Ahmed, Roman A. and Hyndman, Rob J. , date =. Hierarchical Forecasts for. doi:10.1016/j.ijforecast.2008.07.004 , url =

    work page doi:10.1016/j.ijforecast.2008.07.004 2008

  4. [4]

    Bergsteinsson, Hjörleifur and Frölke, Linde and L

    Bacher, Peder and G. Bergsteinsson, Hjörleifur and Frölke, Linde and L. Sørensen, Mikkel and Lemos-Vinasco, Julian and Liisberg, Jon and Kloppenborg Møller, Jan and Aalborg Nielsen, Henrik and Madsen, Henrik , date =. Onlineforecast:. doi:10.32614/RJ-2023-031 , url =

    work page doi:10.32614/rj-2023-031 2023

  5. [5]

    doi:10.1016/j.enbuild.2013.04.022 , url =

    Short-Term Heat Load Forecasting for Single Family Houses , author =. doi:10.1016/j.enbuild.2013.04.022 , url =

    work page doi:10.1016/j.enbuild.2013.04.022 2013

  6. [6]

    Distributed

    Bai, Li and Pinson, Pierre , date =. Distributed. doi:10.3390/en12061112 , url =

    work page doi:10.3390/en12061112

  7. [7]

    Adjoint-

    Bartsch, Jan and Denk, Robert and Volkwein, Stefan , date =. Adjoint-. doi:10.1007/s00245-024-10181-y , url =

    work page doi:10.1007/s00245-024-10181-y

  8. [8]

    doi:10.1016/j.apenergy.2021.116872 , url =

    Heat Load Forecasting Using Adaptive Temporal Hierarchies , author =. doi:10.1016/j.apenergy.2021.116872 , url =

    work page doi:10.1016/j.apenergy.2021.116872 2021

  9. [9]

    doi:10.1016/j.apenergy.2023.121676 , url =

    Heat Load Forecasting Using Adaptive Spatial Hierarchies , author =. doi:10.1016/j.apenergy.2023.121676 , url =

    work page doi:10.1016/j.apenergy.2023.121676 2023

  10. [10]

    doi:10.1080/14786451.2023.2271261 , url =

    Localizing Weather Forecasts for Enhanced Heat Load Forecast Accuracy in Urban District Heating Systems , author =. doi:10.1080/14786451.2023.2271261 , url =

    work page doi:10.1080/14786451.2023.2271261 2023

  11. [11]

    doi:10.1007/978-1-84882-969-5 , url =

    Regression , author =. doi:10.1007/978-1-84882-969-5 , url =

    work page doi:10.1007/978-1-84882-969-5

  12. [12]

    2021 , issn =

    An Introduction to Multivariate Probabilistic Forecast Evaluation , author =. doi:10.1016/j.egyai.2021.100058 , url =

    work page doi:10.1016/j.egyai.2021.100058 2021

  13. [13]

    and Thygesen, Uffe and Bacher, Peder and Madsen, Henrik , date =

    Vetter, Phillip Brinck and Møller, Jan K. and Thygesen, Uffe and Bacher, Peder and Madsen, Henrik , date =

  14. [14]

    Dawid, A P , langid =. Some

  15. [15]

    Fdap.Dk/Teknik/Modeller/ , url =

  16. [16]

    doi:10.1080/0020739830140207 , url =

    An Introduction to Stochastic Differential Equations , author =. doi:10.1080/0020739830140207 , url =

    work page doi:10.1080/0020739830140207

  17. [17]

    doi:10.1016/j.jmva.2018.03.010 , url =

    An Expectation–Maximization Algorithm for the Matrix Normal Distribution with an Application in Remote Sensing , author =. doi:10.1016/j.jmva.2018.03.010 , url =

    work page doi:10.1016/j.jmva.2018.03.010 2018

  18. [18]

    Glanz, Hunter and Carvalho, Luis , date =. An. doi:10.48550/arXiv.1309.6609 , url =. 1309.6609 , eprinttype =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1309.6609

  19. [19]

    doi:10.1002/for.3980090304 , url =

    Disaggregation Methods to Expedite Product Line Forecasting , author =. doi:10.1002/for.3980090304 , url =

    work page doi:10.1002/for.3980090304

  20. [20]

    Matrix Variate Distributions , author =

  21. [21]

    Harris and K

    Harris, Charles R. and Millman, K. Jarrod and family=Walt, given=Stéfan J., prefix=van der, useprefix=true and Gommers, Ralf and Virtanen, Pauli and Cournapeau, David and Wieser, Eric and Taylor, Julian and Berg, Sebastian and Smith, Nathaniel J. and Kern, Robert and Picus, Matti and Hoyer, Stephan and family=Kerkwijk, given=Marten H., prefix=van, usepref...

    work page doi:10.1038/s41586-020-2649-2

  22. [22]

    Hassouna, Mohamed and Holzhüter, Clara and Lytaev, Pawel and Thomas, Josephine and Sick, Bernhard and Scholz, Christoph , date =. Graph. doi:10.1016/j.egyai.2025.100671 , url =. 2407.04522 , eprinttype =

    work page doi:10.1016/j.egyai.2025.100671 2025

  23. [23]

    Optimization with

    Hinze, Michael , date =. Optimization with

  24. [24]

    doi:10.1016/j.csda.2015.11.007 , url =

    Fast Computation of Reconciled Forecasts for Hierarchical and Grouped Time Series , author =. doi:10.1016/j.csda.2015.11.007 , url =

    work page doi:10.1016/j.csda.2015.11.007 2015

  25. [25]

    , author Ahmed, R.A

    Optimal Combination Forecasts for Hierarchical Time Series , author =. doi:10.1016/j.csda.2011.03.006 , url =

    work page doi:10.1016/j.csda.2011.03.006 2011

  26. [26]

    doi:10.1038/s41467-024-49207-y , url =

    Real-Time Outage Management in Active Distribution Networks Using Reinforcement Learning over Graphs , author =. doi:10.1038/s41467-024-49207-y , url =

    work page doi:10.1038/s41467-024-49207-y

  27. [27]

    Modern Trends in Power System Protection for Distribution Grid with High

    Jain, Rishabh and Velaga, Yaswanth Nag and Prabakar, Kumaraguru and Baggu, Murali and Schneider, Kevin , date =. Modern Trends in Power System Protection for Distribution Grid with High. doi:10.1016/j.prime.2022.100080 , url =

    work page doi:10.1016/j.prime.2022.100080 2022

  28. [28]

    doi:10.1016/j.apenergy.2018.05.037 , url =

    Characterizing the Energy Flexibility of Buildings and Districts , author =. doi:10.1016/j.apenergy.2018.05.037 , url =

    work page doi:10.1016/j.apenergy.2018.05.037 2018

  29. [29]

    Stochastic Nonlinear Modelling and Application of Price-Based Energy Flexibility , shorttitle =

    Junker, Rune Grønborg and Kallesøe, Carsten Skovmose and Real, Jaume Palmer and Howard, Bianca and Lopes, Rui Amaral and Madsen, Henrik , date =. Stochastic Nonlinear Modelling and Application of Price-Based Energy Flexibility , shorttitle =. doi:10.1016/j.apenergy.2020.115096 , langid =

    work page doi:10.1016/j.apenergy.2020.115096 2020

  30. [30]

    , date =

    Kalivas, John H. , date =. Overview of Two-Norm (. doi:10.1002/cem.2429 , url =

    work page doi:10.1002/cem.2429

  31. [31]

    Efficient and

    Kidger, Patrick and Foster, James and Li, Xuechen and Lyons, Terry , date =. Efficient and. doi:10.48550/arXiv.2105.13493 , url =. 2105.13493 , eprinttype =

    work page doi:10.48550/arxiv.2105.13493

  32. [32]

    Kidger, Patrick , date =. On. doi:10.48550/arXiv.2202.02435 , url =. 2202.02435 , eprinttype =

    work page doi:10.48550/arxiv.2202.02435

  33. [33]

    Kidger, Patrick and Foster, James and Li, Xuechen and Oberhauser, Harald and Lyons, Terry , langid =. Neural

  34. [34]

    Kingma, Diederik P and Welling, Max , langid =. An

  35. [35]

    and Platen, Eckhard , date =

    Kloeden, Peter E. and Platen, Eckhard , date =. Numerical. doi:10.1007/978-3-662-12616-5 , url =

    work page doi:10.1007/978-3-662-12616-5

  36. [36]

    Cross-Temporal Coherent Forecasts for

    Kourentzes, Nikolaos and Athanasopoulos, George , date =. Cross-Temporal Coherent Forecasts for. doi:10.1016/j.annals.2019.02.001 , url =

    work page doi:10.1016/j.annals.2019.02.001 2019

  37. [37]

    Kristensen, Kasper , date =

  38. [38]

    and Skaug, Hans and Bell, Bradley M

    Kristensen, Kasper and Nielsen, Anders and Berg, Casper W. and Skaug, Hans and Bell, Bradley M. , date =. doi:10.18637/jss.v070.i05 , url =

    work page doi:10.18637/jss.v070.i05

  39. [39]

    doi:10.1016/S0927-5398(03)00007-0 , url =

    Improved Estimation of the Covariance Matrix of Stock Returns with an Application to Portfolio Selection , author =. doi:10.1016/S0927-5398(03)00007-0 , url =

    work page doi:10.1016/s0927-5398(03)00007-0

  40. [40]

    Scalable

    Li, Xuechen and Wong, Ting-Kam Leonard and Chen, Ricky T Q and Duvenaud, David , date =. Scalable

  41. [41]

    Liu, Xuanqing and Xiao, Tesi and Si, Si and Cao, Qin and Kumar, Sanjiv and Hsieh, Cho-Jui , date =. Neural. doi:10.48550/arXiv.1906.02355 , url =. 1906.02355 , eprinttype =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.1906.02355 1906

  42. [42]

    , date =

    Lock, Eric F. , date =. Tensor-on-. doi:10.1080/10618600.2017.1401544 , url =

    work page doi:10.1080/10618600.2017.1401544 2017

  43. [43]

    Introduction to

    Madsen, Henrik and Thyregod, Poul , langid =. Introduction to

  44. [44]

    Modelling

    Madsen, Henrik and Holst, Jan , date =. Modelling

  45. [45]

    Madsen, Henrik and Tohidi, Seyed Shahabaldin and Ebrahimy, Razgar and Banaei, Mohsen and Ritschel, Tobias K. S. and Mahdavi, Nariman , editor =. Recent. Energy. doi:10.1007/978-3-031-74741-0_12 , isbn =

    work page doi:10.1007/978-3-031-74741-0_12

  46. [46]

    Madsen, Henrik , date =. Time

  47. [47]

    and Dutilleul, Pierre , date =

    Manceur, Ameur M. and Dutilleul, Pierre , date =. Maximum Likelihood Estimation for the Tensor Normal Distribution:. doi:10.1016/j.cam.2012.09.017 , url =

    work page doi:10.1016/j.cam.2012.09.017 2012

  48. [48]

    Optimal Price Signal Generation for Indirect Control of Flexible Energy Demand under Uncertainty , author =

  49. [49]

    McKinney, Wes , editor =. Data. Proceedings of the 9th

  50. [50]

    Forecast

    Møller, Jan Kloppenborg and Nystrup, Peter and Hjorth, Poul G and Madsen, Henrik , langid =. Forecast

  51. [51]

    doi:10.1016/j.ijforecast.2022.12.005 , url =

    Likelihood-Based Inference in Temporal Hierarchies , author =. doi:10.1016/j.ijforecast.2022.12.005 , url =

    work page doi:10.1016/j.ijforecast.2022.12.005 2022

  52. [52]

    and Madsen, Henrik , date =

    Møller, Jan Kloppenborg and Nystrup, Peter and Hjorth, Poul G. and Madsen, Henrik , date =. Optimal. doi:10.48550/arXiv.2402.06480 , url =. 2402.06480 , eprinttype =

    work page doi:10.48550/arxiv.2402.06480

  53. [53]

    Statistical

    Møller, Jan Kloppenborg and Schweiker, Marcel and Andersen, Rune Korsholm and Gunay, Burak and Yilmaz, Selin and Barthelmes, Verena Marie and Madsen, Henrik , date =. Statistical. doi:10.1201/9781003340812 , url =

    work page doi:10.1201/9781003340812

  54. [54]

    , date =

    Murphy, Kevin P. , date =. Probabilistic

  55. [55]

    Predicting the

    Nielsen, Henrik Aalborg and Madsen, Henrik , langid =. Predicting the

  56. [56]

    doi:10.1016/j.ijforecast.2020.12.003 , url =

    Dimensionality Reduction in Forecasting with Temporal Hierarchies , author =. doi:10.1016/j.ijforecast.2020.12.003 , url =

    work page doi:10.1016/j.ijforecast.2020.12.003 2020

  57. [57]

    doi:10.1016/j.ejor.2019.07.061 , url =

    Temporal Hierarchies with Autocorrelation for Load Forecasting , author =. doi:10.1016/j.ejor.2019.07.061 , url =

    work page doi:10.1016/j.ejor.2019.07.061 2019

  58. [58]

    Maximum Likelihood Estimation in the Tensor Normal Model with a Structured Mean , author =

  59. [59]

    Stochastic

    Øksendal, Bernt , date =. Stochastic. doi:10.1007/978-3-662-02847-6 , url =

    work page doi:10.1007/978-3-662-02847-6

  60. [60]

    2021 , issn =

    Panagiotelis, Anastasios and Athanasopoulos, George and Gamakumara, Puwasala and Hyndman, Rob J. , date =. Forecast Reconciliation:. doi:10.1016/j.ijforecast.2020.06.004 , url =

    work page doi:10.1016/j.ijforecast.2020.06.004 2020

  61. [61]

    The Matrix Cookbook , author =

  62. [62]

    doi:10.1016/j.ijpe.2021.108221 , url =

    Stochastic Coherency in Forecast Reconciliation , author =. doi:10.1016/j.ijpe.2021.108221 , url =

    work page doi:10.1016/j.ijpe.2021.108221 2021

  63. [63]

    R: A Language and Environment for Statistical Computing , author =

  64. [64]

    doi:10.5281/zenodo.3509134 , url =

    Pandas-Dev/Pandas:. doi:10.5281/zenodo.3509134 , url =

    work page doi:10.5281/zenodo.3509134

  65. [65]

    doi:10.5281/zenodo.20287172 , url =

    Rønlev-Knudsen, Tobias and Madsen, Henrik and Bacher, Peder and Møller, Jan Kloppenborg , date =. doi:10.5281/zenodo.20287172 , url =

    work page doi:10.5281/zenodo.20287172

  66. [66]

    Schäfer, Juliane and Strimmer, Korbinian , date =. A. doi:10.2202/1544-6115.1175 , url =

    work page doi:10.2202/1544-6115.1175

  67. [67]

    doi:10.1016/j.applthermaleng.2025.127770 , url =

    Integration of Pit Thermal Energy Storages into District Heating Networks – a Techno-Economic Case Study , author =. doi:10.1016/j.applthermaleng.2025.127770 , url =

    work page doi:10.1016/j.applthermaleng.2025.127770 2025

  68. [68]

    doi:10.1002/(SICI)1099-128X(199901/02)13:1<31::AID-CEM528>3.0.CO;2-P , url =

    Multiway Covariates Regression Models , author =. doi:10.1002/(SICI)1099-128X(199901/02)13:1<31::AID-CEM528>3.0.CO;2-P , url =

    work page doi:10.1002/(sici)1099-128x(199901/02)13:1

  69. [69]

    Song, Yang and Durkan, Conor and Murray, Iain and Ermon, Stefano , langid =. Maximum

  70. [70]

    Score-Based Generative Modeling through Stochastic Differential Equations

    Song, Yang and Sohl-Dickstein, Jascha and Kingma, Diederik P. and Kumar, Abhishek and Ermon, Stefano and Poole, Ben , date =. Score-. doi:10.48550/arXiv.2011.13456 , url =. 2011.13456 , eprinttype =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2011.13456 2011

  71. [71]

    doi:10.1002/wene.465 , url =

    Recent Developments in Multivariate Wind and Solar Power Forecasting , author =. doi:10.1002/wene.465 , url =

    work page doi:10.1002/wene.465

  72. [72]

    doi:10.1016/j.apenergy.2021.116889 , url =

    Advanced Forecasting and Disturbance Modelling for Model Predictive Control of Smart Energy Systems , author =. doi:10.1016/j.apenergy.2021.116889 , url =

    work page doi:10.1016/j.apenergy.2021.116889 2021

  73. [73]

    Inference in stochastic differential equations using the Laplace approximation: Demonstration and examples

    Thygesen, Uffe Høgsbro and Kristensen, Kasper , date =. Inference in Stochastic Differential Equations Using the. doi:10.48550/arXiv.2503.21358 , url =. 2503.21358 , eprinttype =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2503.21358

  74. [74]

    Thygesen, Uffe Høgsbro , date =

  75. [75]

    Stochastic

    Thygesen, Uffe Høgsbro , date =. Stochastic. doi:10.1201/9781003277569 , isbn =

    work page doi:10.1201/9781003277569

  76. [76]

    Transition Probabilities for Stochastic Differential Equations Using the

    Thygesen, Uffe Høgsbro , date =. Transition Probabilities for Stochastic Differential Equations Using the. doi:10.48550/arXiv.2503.21399 , url =. 2503.21399 , eprinttype =

    work page doi:10.48550/arxiv.2503.21399

  77. [77]

    Validation of Ecological State Space Models Using the

    Thygesen, Uffe Høgsbro and Albertsen, Christoffer Moesgaard and Berg, Casper Willestofte and Kristensen, Kasper and Nielsen, Anders , date =. Validation of Ecological State Space Models Using the. doi:10.1007/s10651-017-0372-4 , url =

    work page doi:10.1007/s10651-017-0372-4

  78. [78]

    1996, Journal of the Royal Statistical Society: Series B (Methodological), 58, 267, doi: 10.1111/j.2517-6161.1996.tb02080.x

    Tibshirani, Robert , date =. Regression. doi:10.1111/j.2517-6161.1996.tb02080.x , url =

    work page doi:10.1111/j.2517-6161.1996.tb02080.x 1996

  79. [79]

    Tohidi, Seyed Shahabaldin and Madsen, Henrik and Tsaousoglou, Georgios and Ritschel, Tobias K. S. , date =. Adaptive. 2024. doi:10.23919/ECC64448.2024.10590768 , url =

    work page doi:10.23919/ecc64448.2024.10590768 2024

  80. [80]

    doi:10.48550/arXiv.2405.15099 , url =

    Stability Analysis of Nonlinear Stochastic Flexibility Function in Smart Energy Systems , author =. doi:10.48550/arXiv.2405.15099 , url =. 2405.15099 , eprinttype =

    work page doi:10.48550/arxiv.2405.15099

Showing first 80 references.