Causal Time-Series Synchronization for Multi-Dimensional Forecasting

Georgios C. Chasparis; Josef K\"ung; Michael Mayr

REVIEW 4 major objections 6 minor 36 references

Causal Time-Series Synchronization for Multi-Dimensional Forecasting

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper claims that shifting cause variables forward by their estimated causal lags improves forecasting accuracy across four channel-dependent models and makes pre-training transferable to unseen causal structures.

desk verdict The shift trick as described cannot put highly lagged causes into the context window, so the paper's central claim is unsupported as written. read the letter →

arxiv 2411.10152 v1 pith:H3FSQB5E submitted 2024-11-15 cs.LG cs.AI

Michael Mayr , Georgios C. Chasparis , Josef Küng This is my paper · ORCID

classification cs.LGcs.AI

keywords time-seriesforecastingcausalsynchronizationGrangercausalitychanneldependencetransferlearningpre-trainingdigitaltwinsprocessindustry

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper claims that channel-dependent forecasting models learn better from temporally aligned cause-effect pairs than from raw multivariate time series. It proposes a data-preparation pipeline: detect lagged cause-effect relationships with a linear Granger causality test (a regression check of whether one series' past helps predict another), shift each cause variable forward by the estimated lag so it lines up with its effect, and construct training windows that keep both the shifted and the unshifted cause alongside the target effect. On synthetic datasets with known causal lags, this synchronization reduces forecast error, measured by mean absolute percentage error (MAPE), for all four tested models, from 7.24 to 5.07 for the weakest baseline and from 3.29 to 2.99 for the strongest. The paper further shows that pre-training on many synchronized source pairs and fine-tuning on a target dataset yields a MAPE of 2.21, the best configuration tested. If the claim holds, it offers a transfer-learning route for process-industry forecasting where variables differ across plants and cause-effect delays can exceed the model's context window.

What carries the argument

The load-bearing object is the target-oriented causal shift: for each detected cause-effect pair $(Z^{(j)}, X^{(i)})$ with estimated lag $\delta_{ij}$, the cause series is shifted forward by $\delta_{ij}$ time steps so that its values align with the effect's present. The shift is produced by a linear VAR Granger causality analysis, a regression-based test of whether past values of one series help predict another, which supplies both the pair selection and the lag. To avoid information loss, the input window stacks the target effect, the shifted cause, and the original unshifted cause, letting the model see both the aligned and the raw series. This construction breaks a complex, possibly disconnected causal graph into many small paired training samples, which the paper argues reduces the noise from causally unrelated or misaligned channels.

What would settle it

Run the same synthetic experiment with oracle lags versus lags perturbed by one to ten time steps; if MAPE degrades to or below the non-synchronized baseline under even small perturbation, the claimed benefit is an artifact of perfect lag estimation. Alternatively, generate data where lags change at known break points; if synchronized training underperforms ordinary multivariate training on non-stationary lags, the global-stationarity assumption fails.

Watch

Extended reading notes

Core claim

The central claim is that the target-oriented causal shift, moving a cause variable's values forward by its estimated lag $\delta_{ij}$ before training, converts a misaligned multivariate forecasting problem into a set of aligned cause-effect subproblems, and that this alignment is what lets channel-dependent models exploit cross-variable structure. The paper calls the resulting training scheme a hybrid CI+CD strategy: within each pair, the model sees the effect and both versions of the cause (shifted and unshifted), while across pairs the same model weights are shared, so the network indirectly learns from many cause-effect relationships without being forced to model causally unrelated channels. Experimentally, the claim is supported by consistent MAPE reductions on four architectures and by the transfer result in Table 2, where synchronized pre-training plus fine-tuning reaches 2.21 MAPE versus 2.56 for the non-synchronized counterpart. The authors state that the improvements are heavily dependent on causal pair extraction and lag estimation, and they assume globally stationary causal lags.

Load-bearing premise

The method stands or falls on the estimated causal lags $\delta_{ij}$ being accurate and stable over time; if a lag is wrong or drifts, shifting the cause aligns it with irrelevant history, and the synchronized training loses its advantage.

Editorial extensions

If this is right

If synchronization works as claimed, channel-dependent pre-training no longer needs a fixed variable set, because each training sample is a self-contained cause-effect pair; this is what makes cross-dataset pre-training possible.
Causes that occurred before the context window become visible after shifting, so models can exploit highly lagged dependencies (hours or days in process plants) without extending the context.
The hybrid CI+CD construction lets a shared-weight model learn cross-variable dependence within pairs and generalizable cause-effect patterns across pairs, which is the basis of the reported transfer gains.
Synchronized pre-training plus fine-tuning outperforms both non-synchronized pre-training and target-only training on the synthetic benchmark, providing a concrete training recipe to test on real industrial data.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The paper's own caveat suggests a direct stress test: if estimated lags are even slightly wrong, the shift places irrelevant values in the context; measuring MAPE against lag-estimation error on the same synthetic generator would show how much of the gain depends on perfect lags.
Because the method decomposes the graph into pairs, it ignores the state of other variables in the causal structure; an extension would be to feed the broader graph context or add a reconciliation loss, which the authors list as future work.
The same pairing and shifting idea could be applied to classification or control tasks in digital twins, not just forecasting, since the aligned cause-effect window is task-agnostic.
If paired pre-training transfers to real process data, it may reduce the need for per-plant labeled data, which is the implicit promise of foundation models for the process industry.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Two things to know about arXiv:2411.10152. The core idea is sensible: identify cause-effect pairs with Granger causality, align them by the estimated lag, and train channel-dependent forecasting models on the aligned pairs rather than on the full multivariate mess. That is a clean preprocessing concept, and the authors deserve credit for making it explicit and for adding the non-synchronized cause channel to avoid discarding information. The second thing is that the shift as implemented does not do what the paper says it does.

The text in §3.2 and Fig. 3 defines the shift as moving the cause forward by δ, i.e., \tilde Z_s = Z_{s-δ}. In a forecasting setting with context [t-C:t] and horizon H, the cause relevant for predicting X_{t+h} is Z_{t+h-δ}. The shifted channel only provides \tilde Z_{t-C:t} = Z_{t-C-δ:t-δ}, which ends at t-δ; the needed value for any h is not there. It appears in the non-synchronized channel only when δ ≤ C+h, which with C=30, H=10 means δ ≤ 40. The paper motivates the whole method with lags up to 200. So for most of the regime it claims to address, neither input channel contains the needed cause. This is not a question of lag-estimation error; it fails even with oracle lags. The stated mechanism cannot achieve the stated goal.

What the paper does well: the pairwise decomposition is a reasonable way to reduce noise from irrelevant channels; the synthetic data with known ground truth is a good test bed; and the empirical improvement in Table 1, if it survives scrutiny, is worth taking seriously. Including the non-synchronized channel is a thoughtful touch. The transfer experiment in Table 2 is a step beyond simple forecasting.

Soft spots beyond the mechanical one: the experiments report no variance or significance tests, so the 9–32% improvements may be within noise. There is no longer-context baseline, which would be the natural ablated control: if a model with C=60 or C=200 captures the same benefit, then the synchronization is irrelevant. The paper never checks how often the estimated Granger lags match the injected ground truth, despite admitting that performance depends heavily on that step. Table 2 actually shows negative transfer for the non-synchronized case (2.56 vs 2.42), which complicates the pre-training claim; only the synchronized pre-training beats supervised training. And no code or data is provided, so reproducibility is weak.

On balance, this is a paper with a real idea but a load-bearing flaw in the presentation of the mechanism, plus an experimental section that needs more rigor. It is honest about many limitations, but the central contradiction is not acknowledged. A serious referee should see it, because the shift issue is fixable in principle (e.g., a longer context or a different shift convention), and the empirical question of whether synchronized pairs help is legitimate. But I would not accept it in its current form.

Recommendation: send it to review, but flag the shift/context issue explicitly. It deserves referee time, not because it is close to correct, but because the failure mode is instructive and the question is worth answering properly.

Referee Report

4 major / 6 minor

Summary. The paper proposes a preprocessing and pre-training strategy for multivariate time-series forecasting in process-industry digital twins. Using linear Granger causality, the method identifies cause-effect pairs and their time lags, then applies a 'target-oriented causal shift' to align each cause series with its effect before training channel-dependent models (TSMixerx, MLPMultivariate, BiTCN, TFT). The shifted cause and a non-shifted copy are fed as inputs alongside the target, and this synchronized format is used both for end-to-end forecasting and for pre-training on multiple synthetic source datasets followed by fine-tuning on a target dataset. Experiments on synthetic data with known injected causal lags report MAPE improvements for all four models in Table 1 and an advantage for synchronized pre-training in Table 2. The central claim is that synchronization brings highly lagged causes into the context window and improves generalization in channel-dependent forecasting.

Significance. If the mechanism worked as stated, the idea of breaking a complex causal structure into synchronized cause-effect pairs for pre-training would be a useful and timely contribution for industrial time-series forecasting, addressing the known weakness of channel-dependent models on misaligned variables. The use of synthetic data with known ground-truth causal lags is a good first validation choice, and including both the synchronized and non-synchronized cause channels is a sensible safeguard against information loss. The paper also tests the strategy across several model families, which strengthens the empirical question under study. However, the central mechanism as written does not deliver the promised out-of-context lag information, and the experimental evidence lacks the uncertainty quantification and baseline controls needed to support the claimed improvements.

major comments (4)

[§3.2, §3.3, Fig. 3] The target-oriented shift as defined does not place the cause values needed for H-step-ahead forecasts into the input context. If shifting means \tilde Z_s = Z_{s-δ_{ij}}, then the synchronized context \tilde Z_{t-C:t} contains cause values Z_{t-C-δ:t-δ}, while the non-synchronized input Z_{t-C:t} contains values up to Z_t. For target step h, X_{t+h} depends on Z_{t+h-δ}; when δ-h > C, that value appears in neither input channel. With the reported C=30, H=10 and lags up to 200, any δ>40 makes the cause for X_{t+H} unavailable to the model. The statement in §4.1 that synchronization shifts 'causes from outside the context window' is therefore not realized by the described pipeline. The authors should either correct the shift definition (for example, a horizon-specific shift of δ-H), add a cause-specific context that extends further back, or show an alternative construction; the experiments must then be rerun under the corrected definition.
[§4.1, Table 1] The empirical core of the paper is a synchronized-versus-non-synchronized comparison on ten synthetic datasets, but Table 1 reports a single MAPE per model with no error bars, no standard deviations, and no statistical test across datasets. The paper also never reports how often the data-driven Granger lag estimates match the injected ground-truth lags, so the reader cannot tell whether the improvement reflects correct synchronization or an artifact of the estimation procedure. The manuscript itself acknowledges in §4.1 that the improvements 'are heavily dependent on causal pair extraction and lag estimation'; that step needs to be validated directly, e.g., by lag-recovery rates and a sensitivity analysis around misspecified δ values.
[§4.2, Table 2] Table 2 does not support the unqualified claim in §4.2 that pre-training on many causal structures improves over supervised training. In the non-synchronized rows, pre-training plus fine-tuning (MAPE 2.56) is worse than supervised training on the target only (MAPE 2.42). Only the synchronized configuration shows the claimed improvement (2.21 vs. 2.38). The claim should be restricted to synchronized pre-training, and the comparison needs per-model results or confidence intervals before a generalization statement can be made.
[§4.1, §2.3] The motivation for synchronization is that highly lagged dependencies may lie outside a fixed context window (§2.3), but no baseline with a larger context length is included. Without training the same models with C larger than the maximum relevant lag (e.g., C > 210), the reported NS-versus-S differences cannot distinguish the proposed synchronization mechanism from a generic reformatting of the inputs. This control is necessary to support the paper's central claim that the method specifically recovers out-of-context causal information.

minor comments (6)

[§4.1, Table 1] The metric is named 'Mean Average Percentage Error' in the text and table caption; the standard name for MAPE is 'Mean Absolute Percentage Error', since the absolute errors are averaged.
[§2.3] The section heading contains the typo 'Channel Indpendence' and should read 'Channel Independence'.
[Fig. 3 caption] The caption says 'context - C denotes the context length, i.e. historic data points'; this should be worded as 'context of length C' to avoid confusion with subtraction.
[§3.1] The threshold used for Granger causality is never specified; please report the p-value threshold or other criterion used to select cause-effect pairs in both experiments.
[§4.2] Experimental details for the transfer experiment are incomplete: the target dataset size, number of variables, composition of the 100 source datasets, and the fine-tuning protocol (layers, learning rates, epochs) are not specified, which limits reproducibility.
[Reference [8]] Reference [8] (Fukushima 1975) does not appear to be the appropriate citation for the 'MLPMultivariate' model used in the experiments; please cite the neuralforecast implementation or another directly relevant source.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: lag estimates are standard supervised inputs, and the reported MAPE gains are empirical held-out outcomes, not identities.

full rationale

The derivation chain is not circular. The causal delay δ_ij is estimated from training data using linear Granger causality (Sec. 3.1) and then used to build the synchronized input \tilde Z_s = Z_{s-δ} (Sec. 3.2). This is standard feature construction: a parameter estimated on data is used to transform inputs, and the forecasting model is evaluated on held-out test samples from a 70/20/10 sliding-window split. The reported MAPE improvements in Tables 1 and 2 are empirical outcomes, not algebraic consequences of the lag estimates. There is no equation in the paper under which the predicted MAPE equals the fitted lag or the synchronization operator by construction. The citation to Ammann et al. [2] is a self-citation because two of the present authors are co-authors, but it is not load-bearing: the paper itself describes the simple linear VAR model and F-test procedure, and the cited work does not contain the synchronization or pre-training result. The shift trick is explicitly credited to Zhao et al. [30] ('similarly to [30]'), so the paper does not relabel a known result as a first-principles derivation. A possible internal-validity concern is that for lags larger than the context window, the shifted cause needed for a future horizon may not actually be present in the input context; that is a correctness issue about whether the mechanism delivers the claimed information, not a circularity, because the improvement claim is empirical rather than a reduction of the output to the input.

Assumptions & free parameters 4 free parameters · 3 assumptions · 0 invented entities

The central claim rests on a small set of estimation choices: the causal lag, the Granger threshold, the maximum lag searched, and the context length. The lags are estimated from data, but the estimation accuracy and sensitivity are not assessed. The paper also relies on explicit domain assumptions of linearity and stationarity, which the authors acknowledge may fail in real systems. No new physical or conceptual entities are introduced.

free parameters (4)

causal_lag_delta_ij = estimated from training data via VAR/Granger, not reported
Each cause-effect pair is shifted by this lag to synchronize the cause with the effect; the method's benefit depends on this estimate, but the values and their accuracy are not reported (Section 3.2, Section 4.1).
granger_threshold = not specified
The threshold for declaring a pair causal is 'expert-defined or data-driven defined' (Section 3.1) but the actual value is not stated.
max_lag_Lmax = not specified
The lag search range for Granger causality is Lmax; the paper does not report its value.
context_length_C = 30 (3 times horizon)
The context window is set to C = H*3 for all experiments, which interacts with the shift: causes with lag greater than C may still be brought into context by the shift.

assumptions (3)

domain assumption Cause-effect relationships are linear and can be modeled with a VAR process.
Section 2.2 and Section 5.1: the paper uses linear Granger causality and notes non-linear dependencies are not addressed.
domain assumption Causal relationships and their lags are globally stationary within each dataset.
Section 3.2: 'The assumption of global static causal lags may not hold in real-world systems'; the shift-based synchronization depends on this assumption.
domain assumption Synthetic data from Tigramite provides ground-truth causal structures that linear Granger can recover.
Section 4.1: the authors generate data with known lags but never report the accuracy of the data-driven lag estimates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Causal Time-Series Synchronization for Multi-Dimensional Forecasting." pith.science (2026). https://pith.science/paper/H3FSQB5E

@misc{pith2026241110152,
  author       = {Pith},
  title        = {Pith review of: Causal Time-Series Synchronization for Multi-Dimensional Forecasting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H3FSQB5E}},
  note         = {Machine review of arXiv:2411.10152}
}

read the original abstract

The process industry's high expectations for Digital Twins require modeling approaches that can generalize across tasks and diverse domains with potentially different data dimensions and distributional shifts i.e., Foundational Models. Despite success in natural language processing and computer vision, transfer learning with (self-) supervised signals for pre-training general-purpose models is largely unexplored in the context of Digital Twins in the process industry due to challenges posed by multi-dimensional time-series data, lagged cause-effect dependencies, complex causal structures, and varying number of (exogenous) variables. We propose a novel channel-dependent pre-training strategy that leverages synchronized cause-effect pairs to overcome these challenges by breaking down the multi-dimensional time-series data into pairs of cause-effect variables. Our approach focuses on: (i) identifying highly lagged causal relationships using data-driven methods, (ii) synchronizing cause-effect pairs to generate training samples for channel-dependent pre-training, and (iii) evaluating the effectiveness of this approach in channel-dependent forecasting. Our experimental results demonstrate significant improvements in forecasting accuracy and generalization capability compared to traditional training methods.

Figures

Figures reproduced from arXiv: 2411.10152 by the authors.

**Figure 2.** Illustrative example of a non-synchronized cause [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Illustrative example of the transformed cause-effe [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 18 canonical work pages

[1]

Abburu, S., Berre, A.J., Jacoby, M., Roman, D., Stojanovic, L., St ojanovic, N.,

work page
[2]

Automated Knowledge Graph Learning in Industrial Processes

Ammann, L., Martinez-Gil, J., Mayr, M., Chasparis, G.C., 2024. Auto mated knowl- edge graph learning in industrial processes. URL: https://arxiv.org/abs/2407.02106, arXiv:2407.02106

work page Pith review arXiv 2024
[3]

Chen, P., Zhang, Y., Cheng, Y., Shu, Y., Wang, Y., Wen, Q., Yang, B., Guo, C.,

work page
[4]

Tsmixer: An all-mlp architecture for time series forecasting

Chen, S.A., Li, C.L., Yoder, N., Arik, S.O., Pﬁster, T., 2023. Tsmixer: An all-mlp architecture for time series forecasting. URL: https://arxiv.org/abs/2303.06053, arXiv:2303.06053. 11

arXiv 2023
[5]

Tiny Time Mixers (TTMs): Fast Pre- trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate T ime Series

Ekambaram, V., Jati, A., Dayama, P., Mukherjee, S., Nguyen, N.H., Giﬀord, W.M., Reddy, C., Kalagnanam, J., 2024. Tiny Time Mixers (TTMs): Fast Pre- trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate T ime Series. URL: http://arxiv.org/abs/2401.03955, doi: 10.48550/arXiv.2401.03955. arXiv:2401.03955 [cs]

arXiv doi:10.48550/arxiv.2401.03955 2024
[6]

Ekambaram, V., Jati, A., Nguyen, N., Sinthong, P., Kalagnanam, J., 2023. TSMixer: Lightweight MLP-Mixer Model for Multivariate Time Series Forecastin g, in: Proceed- ings of the 29th ACM SIGKDD Conference on Knowledge Discovery an d Data Min- ing, Association for Computing Machinery, New York, NY, USA. pp. 4 59–469. URL: https://doi.org/10.1145/3580305.35...

arXiv 2023
[7]

Making the eu climate-neutral by 2050

EU, . Making the eu climate-neutral by 2050. URL: https://ec.europa.eu/commission/presscorner/detail/en/ip_20_335

work page
[8]

SELF-ORGANIZING MULTILAYERED NEURA L NETWORK

FUKUSHIMA, K., 1975. SELF-ORGANIZING MULTILAYERED NEURA L NETWORK. SELF-ORGANIZING MULTILAYERED NEURAL NETWORK

work page 1975

Show all 36 references

[9]

TimeGPT- 1

Garza, A., Challu, C., Mergenthaler-Canseco, M., 2024. TimeGPT- 1. URL: http://arxiv.org/abs/2310.03589, doi: 10.48550/arXiv.2310.03589. arXiv:2310.03589 [cs, stat]

2024 doi arXiv
[10]

Investigating Causal Relations by Econo metric Models and Cross-spectral Methods

Granger, C.W.J., 1969. Investigating Causal Relations by Econo metric Models and Cross-spectral Methods. Econometrica 37, 424–438 . URL: https://www.jstor.org/stable/1912791, doi: 10.2307/1912791. publisher: [Wiley, Econometric Society]

1969
[11]

CrossGNN: Confronting Noisy Multivariate Time Series Via Cross Inte raction Reﬁne- ment

Huang, Q., Shen, L., Zhang, R., Ding, S., Wang, B., Zhou, Z., Wang, Y., 2023. CrossGNN: Confronting Noisy Multivariate Time Series Via Cross Inte raction Reﬁne- ment. Advances in Neural Information Processing Systems 36, 46 885–46902. URL: https://proceedings.neurips.cc/paper_...

2023
[12]

Digital Twin in manu- facturing: A categorical literature review and classiﬁcation

Kritzinger, W., Karner, M., Traar, G., Henjes, J., Sihn, W., 2018. Digital Twin in manu- facturing: A categorical literature review and classiﬁcation. IF AC -PapersOnLine 51, 1016–

2018
[13]

In dustry 4.0

Lasi, H., Fettke, P., Kemper, H.G., Feld, T., Hoﬀmann, M., 2014. In dustry 4.0. Bus Inf Syst Eng 6, 239–242. URL: https://doi.org/10.1007/s12599-014-0334-4 , doi:10.1007/s12599-014-0334-4

2014 doi
[14]

Learning to Embed Time Series Pa tches Inde- pendently

Lee, S., Park, T., Lee, K., 2024. Learning to Embed Time Series Pa tches Inde- pendently. URL: http://arxiv.org/abs/2312.16427, doi: 10.48550/arXiv.2312.16427. arXiv:2312.16427 [cs, stat]

2024 doi arXiv
[15]

Temporal fusion tr ansformers for inter- pretable multi-horizon time series forecasting

Lim, B., Arik, S.O., Loeﬀ, N., Pﬁster, T., 2020. Temporal fusion tr ansformers for inter- pretable multi-horizon time series forecasting. URL: https://arxiv.org/abs/1912.09363, arXiv:1912.09363. 12

2020 arXiv
[16]

iT rans- former: Inverted Transformers Are Eﬀective for Time Series For ecasting

Liu, Y., Hu, T., Zhang, H., Wu, H., Wang, S., Ma, L., Long, M., 2024. iT rans- former: Inverted Transformers Are Eﬀective for Time Series For ecasting. URL: http://arxiv.org/abs/2310.06625, doi: 10.48550/arXiv.2310.06625. arXiv:2310.06625 [cs]

2024 doi arXiv
[17]

Learning Paradigms an d Modelling Methodologies for Digital Twins in Process Industry, pp

Mayr, M., Chasparis, G., K¨ ung, J., 2024. Learning Paradigms an d Modelling Methodologies for Digital Twins in Process Industry, pp. 34–47. doi: 10.1007/978-3-031-68323-7_3

2024 doi
[18]

Abstracting Process Mining Event Logs From Process-State Data To Monitor Control-Flow Of Industrial Manuf acturing Processes

Mayr, M., Luftensteiner, S., Chasparis, G., 2022. Abstracting Process Mining Event Logs From Process-State Data To Monitor Control-Flow Of Industrial Manuf acturing Processes. Procedia Computer Science 200, 1442–1450. doi: 10.1016/j.procs.2022.01.345

2022 doi
[19]

BasisFormer: Attention- based Time Series Forecasting with Learnable and Interpretable Ba sis

Ni, Z., Yu, H., Liu, S., Li, J., Lin, W., 2023. BasisFormer: Attention- based Time Series Forecasting with Learnable and Interpretable Ba sis. Ad- vances in Neural Information Processing Systems 36, 71222–712 41. URL: https://proceedings.neurips.cc/paper_files/paper/2023/hash/...

2023
[20]

A Time Se ries is Worth 64 Words: Long-term Forecasting with Transformers

Nie, Y., Nguyen, N.H., Sinthong, P., Kalagnanam, J., 2023. A Time Se ries is Worth 64 Words: Long-term Forecasting with Transformers. URL: http://arxiv.org/abs/2211.14730, doi:10.48550/arXiv.2211.14730. arXiv:2211.14730 [cs]

2023 doi arXiv
[21]

NeuralForecast: User friendly state-of-the-art neural forecasting models

Olivares, K.G., Chall´ u, C., Garza, F., Canseco, M.M., Dubrawski, A ., 2022. NeuralForecast: User friendly state-of-the-art neural forecasting models. PyC on Salt Lake City, Utah, US 2022. URL: https://github.com/Nixtla/neuralforecast

2022
[22]

Digital Twin: Values, Ch allenges and Enablers From a Modeling Perspective

Rasheed, A., San, O., Kvamsdal, T., 2020. Digital Twin: Values, Ch allenges and Enablers From a Modeling Perspective. IEEE Access 8, 21980–2 2012. URL: https://ieeexplore.ieee.org/document/8972429, doi: 10.1109/ACCESS.2020.2970143. conference Name: IEEE Access

2020
[23]

Lag-Llama: Towards Foundation Models for Probabilistic Time Series F orecasting

Rasul, K., Ashok, A., Williams, A.R., Ghonia, H., Bhagwatkar, R., Khor asani, A., Bayazi, M.J.D., Adamopoulos, G., Riachi, R., Hassen, N., Biloˇ s, M., Garg, S., Schnei- der, A., Chapados, N., Drouin, A., Zantedeschi, V., Nevmyvaka, Y., R ish, I., 2024. Lag-Llama: Towards Foun...

2024 doi arXiv
[24]

Runge, J., Gerhardus, A., Varando, G., Eyring, V., Camps-Valls, G.,
[25]

Parameter eﬃc ient deep probabilistic forecast- ing

Sprangers, O., Schelter, S., de Rijke, M., 2021. Parameter eﬃc ient deep probabilistic forecast- ing. URL: https://arxiv.org/abs/2112.02905, arXiv:2112.02905

2021 arXiv
[26]

Wang, S., Wu, H., Shi, X., Hu, T., Luo, H., Ma, L., Zhang, J.Y., Zhou, J .,
[27]

Uniﬁed Training of Universal Time Series Forecasting Transformers

Woo, G., Liu, C., Kumar, A., Xiong, C., Savarese, S., Sahoo, D., 202 4. Uniﬁed Training of Universal Time Series Forecasting Transformers. URL: http://arxiv.org/abs/2402.02592, doi:10.48550/arXiv.2402.02592. arXiv:2402.02592 [cs]

doi arXiv
[28]

Times- Net: Temporal 2D-Variation Modeling for General Time Series Analys is

Wu, H., Hu, T., Liu, Y., Zhou, H., Wang, J., Long, M., 2023. Times- Net: Temporal 2D-Variation Modeling for General Time Series Analys is. URL: http://arxiv.org/abs/2210.02186, doi: 10.48550/arXiv.2210.02186. arXiv:2210.02186 [cs]

2023 doi arXiv
[29]

Yi, K., Zhang, Q., Fan, W., He, H., Hu, L., Wang, P., An, N., Cao, L., Niu , Z.,
[30]

Rethinking Channel Dependence for M ultivariate Time Series Forecasting: Learning from Leading Indicators

Zhao, L., Shen, Y., 2024. Rethinking Channel Dependence for M ultivariate Time Series Forecasting: Learning from Leading Indicators. URL: http://arxiv.org/abs/2401.17548, doi:10.48550/arXiv.2401.17548. arXiv:2401.17548 [cs]. 14

2024 doi arXiv
[31]

URL: http://arxiv.org/abs/2405.14616, doi: 10.48550/arXiv.2405.14616

TimeMixer: Decomposable Multiscale Mixing for Time Series Forec asting. URL: http://arxiv.org/abs/2405.14616, doi: 10.48550/arXiv.2405.14616. arXiv:2405.14616 [cs]. 13

doi arXiv
[35]

Advances in Neural Information Processing System s 36, 69638–69660

FourierGNN: Rethinking Multivariate Time Series Forecasting f rom a Pure Graph Perspective. Advances in Neural Information Processing System s 36, 69638–69660. URL: https://proceedings.neurips.cc/paper_files/paper/2023/hash/dc1e32dd3eb381dbc71482f6a96cbf86-Abst

2023
[1022]

URL: https://www.sciencedirect.com/science/article/pii/S2405896318316021, doi:https://doi.org/10.1016/j.ifacol.2018.08.474

2018 doi
[2020]

COGNITWIN – Hybrid and Cognitive Digital Twins for the Proces s Indus- try, in: 2020 IEEE International Conference on Engineering, Tec hnology and Innova- tion (ICE/ITMC), pp. 1–8. URL: https://ieeexplore.ieee.org/document/9198403, doi:10.1109/ICE/ITMC49519.2020.9198403

2020
[2023]

Nat Rev Earth Environ 4, 487–505

Causal inference for time series. Nat Rev Earth Environ 4, 487–505. URL: https://www.nature.com/articles/s43017-023-00431-y , doi:10.1038/s43017-023-00431-y

doi
[2024]

URL: http://arxiv.org/abs/2402.05956, doi: 10.48550/arXiv.2402.05956

Pathformer: Multi-scale Transformers with Adaptive Pathw ays for Time Series Forecasting. URL: http://arxiv.org/abs/2402.05956, doi: 10.48550/arXiv.2402.05956. arXiv:2402.05956 [cs]

doi arXiv

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Abburu, S., Berre, A.J., Jacoby, M., Roman, D., Stojanovic, L., St ojanovic, N.,

work page

[2] [2]

Automated Knowledge Graph Learning in Industrial Processes

Ammann, L., Martinez-Gil, J., Mayr, M., Chasparis, G.C., 2024. Auto mated knowl- edge graph learning in industrial processes. URL: https://arxiv.org/abs/2407.02106, arXiv:2407.02106

work page Pith review arXiv 2024

[3] [3]

Chen, P., Zhang, Y., Cheng, Y., Shu, Y., Wang, Y., Wen, Q., Yang, B., Guo, C.,

work page

[4] [4]

Tsmixer: An all-mlp architecture for time series forecasting

Chen, S.A., Li, C.L., Yoder, N., Arik, S.O., Pﬁster, T., 2023. Tsmixer: An all-mlp architecture for time series forecasting. URL: https://arxiv.org/abs/2303.06053, arXiv:2303.06053. 11

arXiv 2023

[5] [5]

Tiny Time Mixers (TTMs): Fast Pre- trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate T ime Series

Ekambaram, V., Jati, A., Dayama, P., Mukherjee, S., Nguyen, N.H., Giﬀord, W.M., Reddy, C., Kalagnanam, J., 2024. Tiny Time Mixers (TTMs): Fast Pre- trained Models for Enhanced Zero/Few-Shot Forecasting of Multivariate T ime Series. URL: http://arxiv.org/abs/2401.03955, doi: 10.48550/arXiv.2401.03955. arXiv:2401.03955 [cs]

arXiv doi:10.48550/arxiv.2401.03955 2024

[6] [6]

Ekambaram, V., Jati, A., Nguyen, N., Sinthong, P., Kalagnanam, J., 2023. TSMixer: Lightweight MLP-Mixer Model for Multivariate Time Series Forecastin g, in: Proceed- ings of the 29th ACM SIGKDD Conference on Knowledge Discovery an d Data Min- ing, Association for Computing Machinery, New York, NY, USA. pp. 4 59–469. URL: https://doi.org/10.1145/3580305.35...

arXiv 2023

[7] [7]

Making the eu climate-neutral by 2050

EU, . Making the eu climate-neutral by 2050. URL: https://ec.europa.eu/commission/presscorner/detail/en/ip_20_335

work page

[8] [8]

SELF-ORGANIZING MULTILAYERED NEURA L NETWORK

FUKUSHIMA, K., 1975. SELF-ORGANIZING MULTILAYERED NEURA L NETWORK. SELF-ORGANIZING MULTILAYERED NEURAL NETWORK

work page 1975