Pith. sign in

REVIEW 3 major objections 5 minor 84 references

Causal View of Time Series Imputation: Some Identification Results on Missing Mechanism

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read This paper claims that time series imputation under MAR and MNAR missingness can recover both the temporal latent states and the latent missing-cause variables, and that these recoveries are component-wise identifiable.

desk verdict The paper's new missing-cause identifiability claim is asserted rather than proven — c_t is never connected to observed data — but the empirical study is extensive and the method may still be useful. read the letter →

arxiv 2505.07180 v1 pith:MIJOEARY submitted 2025-05-12 cs.LG stat.ML

Ruichu Cai , Kaitao Zheng , Junxian Huang , Zijian Li , Zhengming Chen , Boyan Xu , Zhifeng Hao This is my paper · ORCID
classification cs.LGstat.ML
keywords timeseriesimputationmissingatrandomnotidentifiabilitynonlinearindependentcomponentanalysisvariationalinferencenormalizingflowm-graph
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 argues that imputation models fail when they ignore which missing-data mechanism generated the gaps. It proposes DMM, a variational framework with separate architectures for missing-at-random (MAR) and missing-not-at-random (MNAR) data, in which temporal latent states $z_t$ and missing-cause variables $c_t$ are estimated jointly. The central theoretical claim is that, under smooth strictly positive conditionally factorized densities and a linear-independence condition, $z_t$ and $c_t$ are component-wise identifiable under both MAR and MNAR. If true, imputation can be tailored to the actual mechanism and the missingness itself becomes interpretable. Experiments on semisynthetic and benchmark time series show that mechanism-matched DMM models reduce imputation error relative to single-mechanism baselines.

What carries the argument

The central object is the imputation m-graph, a causal graph with observed variables $x^o_t$, missing variables $x^m_t$, latent temporal states $z_t$, and latent missing-cause variables $c_t$. The argument runs on a Jacobian-sparsity mechanism: matching the observed marginal forces an invertible map $h$ between estimated and true latents, conditional independence makes mixed second cross-derivatives of the log-density vanish, and the linear-independence assumptions A2/A3 imply that each row of the Jacobian has exactly one non-zero entry, giving component-wise identifiability. A normalizing-flow prior estimator supplies the log-probabilities and change-of-variable Jacobians needed to compute the KL terms for both $z_t$ and $c_t$.

What would settle it

Simulate data with known ground-truth $c_t$ that generates the mask, train DMM with the mask withheld, and measure the correlation between estimated and true $c_t$; if $c_t$ is not recovered while $z_t$ is, the identifiability claim is false. A sharper version constructs two different ground-truth data-generation processes with the same observed marginal but different $c_t$, and asks the model to distinguish them.

Watch

Extended reading notes

Core claim

The paper establishes identifiability theorems for the time series imputation setting: if observations follow $x_t = g(z_t)$ with an invertible nonlinear mixing $g$, the temporal latent process $z_{t,i}=f_i(z_{t-\tau}, ...)$, and latent missing causes $c_t$ that drive $x^m_t$, then the learned latent states and missing causes are recoverable up to a component-wise invertible transformation. The proof starts from matched marginal $p(\hat{x}_t)=p(x_t)$, derives $p(h(\hat{z}_t))=p(z_t)$ for the invertible map $h = g^{-1} \circ \hat{g}$, uses conditional independence of latent components to force cross-derivatives of the log-density to zero, and applies the linear-independence conditions to conclude that the Jacobian of $h$ has a single non-zero entry per row. The same structure, conditioned on $x^o_t$ for MAR and on $x_{t-1}$ for MNAR, yields component-wise identifiability of $c_t$.

Load-bearing premise

The identifiability of the missing causes rests on the assumption that $c_t$ can be recovered from the observed time series alone, even though the model never includes the pattern of which values are absent that $c_t$ is supposed to generate.

Editorial extensions

If this is right

  • A practitioner can run the mechanism-matched DMM, read off $z_t$ and $c_t$, and impute by sampling from $p(x^m_t \mid z_t, c_t, x^o_t)$ rather than using a one-size-fits-all model.
  • Model selection between DMM-MAR and DMM-MNAR serves as a missing-mechanism test: the variant with better validation performance indicates the likely mechanism, and the experiments show mismatching the mechanism degrades results.
  • MCAR data is declared not identifiable without auxiliary variables, so the framework predicts that latent-cause recovery cannot explain imputation performance on MCAR-dominated data.
  • The recovered missing-cause variables provide an interpretable reason for why values are absent, not only a numerical filling of the gaps.

Reading between the lines

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

  • The model-selection comparison could be turned into an explicit missing-mechanism classifier: train DMM-MAR and DMM-MNAR on labeled synthetic data, then use relative ELBO or imputation error on a new series to announce the mechanism.
  • The same identification strategy plausibly transfers to other mask-driven generative models, such as image inpainting or tabular missing-value recovery, whenever the conditional-independence and linear-independence conditions hold.
  • In clinical time series, an identifiable missing-cause variable could become a decision-support feature, signaling patient disengagement before the next scheduled visit.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes DMM, a variational framework for time series imputation that treats MAR and MNAR missing mechanisms separately. It models data generation through temporal latent states z_t and missing-cause variables c_t, estimates their priors with normalizing-flow architectures, and claims component-wise identifiability of both z_t and c_t under nonlinear ICA assumptions (Theorems 1 and 2). The authors support the framework with extensive experiments on simulated and real-world datasets, including ablation, sensitivity, mixed-mechanism, and healthcare (MIMIC-III) evaluations.

Significance. If the central identifiability claim were valid, the paper would make a useful contribution by linking missing-data mechanisms to causal representation learning and providing mechanism-specific imputation guarantees. The empirical study is broad and includes useful checks (simulation MCC, sensitivity analysis, ablation, mixed mechanisms). However, the claimed identifiability of the missing-cause variable c_t is not established: the proof simply asserts a distributional equality for c_t without deriving it from any observed marginal, and the missingness mask, which is the natural observed consequence of c_t, is absent from the model. This gap affects the main theoretical contribution of the paper. The z_t part follows the standard nonlinear-ICA argument of Yao et al. (2022) and is likely sound under the stated assumptions, but the c_t part is not. On balance, the results as presented do not support the advertised guarantees.

major comments (3)
  1. [Section 4, Appendix C] The proof of Theorem 1 (Appendix C, after Eq. (30)) begins the c_t argument with the assertion 'Similarly, we have p(h_c(\hat c_t)) = p(c_t)', and Theorem 2 (after Eq. (41)) repeats this for p(\hat c_t|x_{t-1}). This equality is not derived from any observed marginal. For z_t, the proof starts from the matched marginal p(\hat x_t) = p(x_t); for c_t, the only variable it is claimed to generate, x^m_t, is missing by definition, and the missingness mask is absent from the generative factorizations (Eqs. (3)-(4)) and from the ELBOs (Eqs. (5) and (13)). Consequently, an arbitrary smooth invertible reparameterization c'_t = h(c_t), with the prior p(c'_t|x^o_t) adjusted accordingly, leaves every term in the training objective unchanged. The linear independence condition in A2 (and A3) is stated in terms of derivatives of p(c_t|x^o_t), which is a quantity involving an unobserved variable; the condition cannot be checked or exploited without an observed variable that depends on c_t. Component-wise identifiability of c_t therefore does not follow from the stated proof. This is load-bearing because the abstract and Section 4 claim identifiability of both z_t and c_t.
  2. [Eqs. (5), (13); Figure 1] The imputation m-graph in Figure 1(a,b) includes c_t -> x^m_t, but x^m_t is the missing part of the data. The standard missing-data machinery uses the mask indicator as the observed consequence of the missingness mechanism; here the mask is used to create training data (Section B.2) but never appears as a modeled random variable in the factorizations or the ELBOs. To identify c_t, one would need a likelihood term p(mask_t | c_t, ...) or p(mask_t | x^o_t, x^m_t) and a corresponding observed marginal to match. The ELBOs contain only the reconstruction term p(x^m_{1:T}|z_{1:T},c_{1:T}) (evaluated on partially observed or imputed targets) and KL terms for z and c. Since c_t has no observed descendant in the model, the observed-data likelihood is invariant to reparameterizations of c_t. The paper never states which observed quantity identifies p(c_t|x^o_t) in Eq. (3) or p(c_t|x_{t-1}) in Eq. (4).
  3. [Appendix F.2, Eqs. (63)-(64)] There is an inconsistency between the generative process used for identification and the ELBO actually optimized. The identification proof for c_t in Theorem 1 targets p(c_t|x^o_t), and Theorem 2 targets p(c_t|x_{t-1}); these conditional distributions appear in the joint factorizations (3) and (4). However, the ELBO derivation in Appendix F.2 (Eq. (63) for MAR and Eq. (64) for MNAR) uses the factorization p(c_{1:T}|z_{1:T}) in the generative model, and Eq. (5)/(13) in the main text use a KL divergence against p(c_{1:T}) without the conditioning on x^o_t or x_{t-1}. The normalizing-flow prior estimators in Sections 3.2 and 3.4 estimate p(\hat c_t|x^o_t) (or p(\hat c_t|x_{t-1})), which are then plugged into the KL terms. This mixing of p(c|x), p(c|z), and p(c) means the optimized objective does not correspond to the generative process for which identifiability is claimed, and it is unclear which distribution is being identified.
minor comments (5)
  1. [Sections 3.2 and 3.4] The symbol p(\hat c_t|x^o_t) is used both as a prior and as a variational posterior; Eqs. (11)-(12) and (19)-(20) should clarify whether the normalizing flow estimates the prior or the approximation q.
  2. [Tables 3 and 12] Table 3 and Table 12 report MCC for z and c on simulated data, but there is no comparison to a trivial baseline (e.g., predicting c_t from x^o_t without any missingness) and no confidence intervals; the claim of 'high estimation accuracy' is not statistically supported.
  3. [Eq. (13)] The notation in Eq. (13) writes E_{q(z_{1:T},c_{1:T}|x_{1:T})} but the second and third terms use q(z|x^o) and q(c|x_{1:T-1}); the joint q is not consistently defined.
  4. [Abstract and Section 4] The abstract says 'latent variables are identifiable' without qualification; the theorems give sufficient conditions for component-wise identifiability up to an invertible transformation, which should be stated explicitly in the abstract.
  5. [Throughout] There are numerous typos and awkward phrasings (e.g., 'one of the most challenge problems', 'the causal relationship between observed and missing variables are unknown', 'we can with the RHS of Equation (47)'), which should be corrected before any resubmission.

Circularity Check

1 steps flagged · score 6.0 of 10

The c_t identifiability proof asserts the target equality p(h_c(\hat c_t)) = p(c_t) without deriving it from observed marginals; the z_t proof remains independent.

  1. self definitional [Appendix C, proof of Theorem 1, Eq. (31); repeated in proof of Theorem 2, Eq. (42)]
    "Similarly, we have p(hc(ˆct)) = p(ct), then logp( ˆct|xo t ) = logp(c t|xo t ) + log|JA hc|= nX k=1 logp(ct,k|xo t ) + log|JA hc| (31)"

    The theorem defines c_t as identifiable when there exists an invertible h_c with c_t = h_c(ˆc_t), which is equivalent to p(h_c(ˆc_t)) = p(c_t). The proof opens the c_t part by asserting exactly this equality. For z_t, the analogous equality is derived from a matched observed marginal, p(ˆx_t)=p(x_t); for c_t no such marginal is available. In Eqs. (3)-(4) c_t appears only as a parent of the missing block x^m_t, and the ELBOs in Eqs. (5) and (13) contain no term that ties c_t to the observed mask or to any observed statistic. The equality p(h_c(ˆc_t)) = p(c_t) is therefore not derived from the data-generation process; it is the target identification statement imported as a premise. The same unsupported assertion is repeated for MNAR in Eq. (42).

full rationale

The paper's central theoretical claim is component-wise identifiability of both temporal latent states z_t and missing-cause variables c_t under MAR and MNAR. The z_t proof is a standard nonlinear-ICA/TDRL argument: it starts from the matched marginal p(ˆx_t)=p(x_t), derives the transformation h, and uses the linear-independence assumptions to conclude component-wise identifiability. That portion is self-contained and not circular; it also borrows standard assumptions from external prior work rather than from the present authors. The c_t proof, however, is circular in a specific, quotable way. In Appendix C the proof simply states 'Similarly, we have p(h_c(ˆc_t)) = p(c_t)' and then differentiates this assumed equality. No observed marginal involving c_t is matched first. The generative factorizations in Eqs. (3) and (4) place c_t only as a cause of the missing block x^m_t, and the ELBOs in Eqs. (5) and (13) contain no modeled factor for the missingness mask and no term in which an observed quantity is equated to a function of c_t. Thus the statement p(h_c(ˆc_t)) = p(c_t) is exactly the identifiability conclusion being proved, making the c_t part of Theorems 1 and 2 reduce by construction to its own conclusion. The simulation experiments, including the MCC results for c_t in Table 12, provide empirical evidence that the method can recover c_t in specific generative settings, and the imputation benchmarks are external comparisons, so the circularity is partial rather than total. The overall score is 6: the central c_t identification guarantee is partially circular, while the z_t identification and the empirical contributions retain independent content.

Assumptions & free parameters 2 free parameters · 5 assumptions · 1 invented entities

The model rests on standard nonlinear-ICA assumptions inherited from TDRL plus the new, unsupported assumption that the missing-cause latent c_t can be identified from the imputation objective without modeling the mask. This determines the central claim's fragility.

free parameters (2)
  • beta (KL weight on latent state prior) = not reported; sensitivity grid 0, 1e-4, 1e-3, 1e-2, 0.1 in Table 14
    Coefficient of L_z^K in Equation (21); tuned and not fixed in main experiments.
  • gamma (KL weight on missing cause prior) = not reported; same sensitivity grid
    Coefficient of L_c^K in Equation (21); tuned and not fixed in main experiments.
assumptions (5)
  • domain assumption A1: p(z_t|z_{t-1}) > 0 and p(c_t|x^o_t) > 0 are smooth and conditionally independent component-wise
    Assumption in Theorems 1 and 2, inherited from Yao et al. 2022; needed for the derivative equations in the proof.
  • domain assumption A2/A3: linear independence of the derivative vectors v and w for z and c
    Assumption in Theorems 1 and 2; ensures the system of equations in the proof has a unique solution.
  • domain assumption x_t = g(z_t) with invertible nonlinear g
    Equation (1); the proof uses invertibility of g and g_hat to relate marginals; the paper itself lists this as a limitation in Appendix G.
  • domain assumption Latent process is first-order Markov with independent noise epsilon_t
    Equation (2) and Appendix F; used to derive change-of-variables priors in Equations (8)-(12).
  • ad hoc to paper The generative factorization includes c_t and omits the mask indicator
    Equations (3)-(4) define the joint distribution with c_t but no term for the observed mask; this is the key unsupported modeling choice.
invented entities (1)
  • missing cause variable c_t
    purpose: Latent variable intended to represent the mechanism that causes entries to be missing under MAR and MNAR
    The mask is the observable consequence of c_t, but the model never includes a likelihood for the mask; the only evaluation against ground truth is on simulated data, with no real-world handle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Causal View of Time Series Imputation: Some Identification Results on Missing Mechanism." pith.science (2026). https://pith.science/paper/MIJOEARY

@misc{pith2026250507180,
  author       = {Pith},
  title        = {Pith review of: Causal View of Time Series Imputation: Some Identification Results on Missing Mechanism},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MIJOEARY}},
  note         = {Machine review of arXiv:2505.07180}
}
read the original abstract

Time series imputation is one of the most challenge problems and has broad applications in various fields like health care and the Internet of Things. Existing methods mainly aim to model the temporally latent dependencies and the generation process from the observed time series data. In real-world scenarios, different types of missing mechanisms, like MAR (Missing At Random), and MNAR (Missing Not At Random) can occur in time series data. However, existing methods often overlook the difference among the aforementioned missing mechanisms and use a single model for time series imputation, which can easily lead to misleading results due to mechanism mismatching. In this paper, we propose a framework for time series imputation problem by exploring Different Missing Mechanisms (DMM in short) and tailoring solutions accordingly. Specifically, we first analyze the data generation processes with temporal latent states and missing cause variables for different mechanisms. Sequentially, we model these generation processes via variational inference and estimate prior distributions of latent variables via normalizing flow-based neural architecture. Furthermore, we establish identifiability results under the nonlinear independent component analysis framework to show that latent variables are identifiable. Experimental results show that our method surpasses existing time series imputation techniques across various datasets with different missing mechanisms, demonstrating its effectiveness in real-world applications.

Figures

Figures reproduced from arXiv: 2505.07180 by the authors.

Figure 1
Figure 1. Data generation processes of time series data under different missing mechanisms. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the DMM framework. X o are the observed variables, X m are the missing data. The latent state variables z1:T and the missing cause variables c1:T are extracted from the encoder. The latent state and missing cause prior networks for DMM-MAR and DMM-MNAR are used to estimate the prior distributions. model to cover all the missing mechanisms, making it hard to identify the distribution P(Xo , Xm). In ge… view at source ↗
Figure 3
Figure 3. Ablation study on the ETTh1 datasets. Data Preprocessing. To facilitate the training of the dataset, we introduce a mask matrix R, to denote the missing positions in the time series data. Consequently, the representation of the time series data, as defined by R, is as follows: X = R ∗ Xo + (1 − R) ∗ Xm (23) Depending on the two distinct data missing mechanisms, we can generate various mask matrices R. Generation of … view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Computational efficiency of ETTh2 dataset with a missing rate of 0.6 under unsupervised conditions in MNAR [PITH_FULL_IMAGE:figures/full_fig_p015_4.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

84 extracted references · 60 canonical work pages

  1. [1]

    The treatment of missing values and its ef- fect on classifier accuracy

    [Acuna and Rodriguez, 2004] Edgar Acuna and Caroline Rodriguez. The treatment of missing values and its ef- fect on classifier accuracy. InClassification, Clustering, and Data Mining Applications: Proceedings of the Meet- ing of the International Federation of Classification So- cieties (IFCS), Illinois Institute of Technology, Chicago, 15–18 July 2004, p...

    2004

  2. [2]

    From an application perspective, these assumptions rigorously defined a subset of applicable scenarios of our model

    at the same time, even if some of the above assumptions do not hold, our method may still learn the correct model. From an application perspective, these assumptions rigorously defined a subset of applicable scenarios of our model. Thus, we provide detailed explanations of the assumptions, how they relate to real-world scenarios, and in which scenarios th...

    work page 2022

  3. [4]

    MIMIC-III5 [Johnsonet al., 2016 ] is a published dataset with de-identified health-related data associated with more than forty thousand patients who stayed in critical care units of the Beth Israel Deaconess Medical Center between 2001 and

    work page 2016

  4. [5]

    Brits: Bidirectional recurrent im- putation for time series.Advances in neural information processing systems, 31,

    [Caoet al., 2018 ] Wei Cao, Dong Wang, Jian Li, Hao Zhou, Lei Li, and Yitan Li. Brits: Bidirectional recurrent im- putation for time series.Advances in neural information processing systems, 31,

    2018

  5. [6]

    Compared with other models for time-series imputation, we can find that the proposed DMM has the best model performance and relatively good model efficiency. Figure 4: Computational efficiency of ETTh2 dataset with a missing rate of 0.6 under unsupervised conditions in MNAR C Proof Theorem 1.(Identification of Latent States and Missing Causes under MAR.)S...

    work page arXiv 2022

  6. [7]

    Provably convergent schr ¨odinger bridge with appli- cations to probabilistic time series imputation

    [Chenet al., 2023 ] Yu Chen, Wei Deng, Shikai Fang, Feng- pei Li, Nicole Tianjiao Yang, Yikai Zhang, Kashif Rasul, Shandian Zhe, Anderson Schneider, and Yuriy Nevmy- vaka. Provably convergent schr ¨odinger bridge with appli- cations to probabilistic time series imputation. InInter- national Conference on Machine Learning, pages 4485–

    work page 2023

  7. [10]

    Filling the g ap s: Multivariate time series imputation by graph neural networks.arXiv preprint arXiv:2108.00298,

    [Ciniet al., 2021 ] Andrea Cini, Ivan Marisca, and Cesare Alippi. Filling the g ap s: Multivariate time series imputation by graph neural networks.arXiv preprint arXiv:2108.00298,

    arXiv 2021

  8. [13]

    Maximum likelihood from incom- plete data via the em algorithm.Journal of the royal statis- tical society: series B (methodological), 39(1):1–22,

    [Dempsteret al., 1977 ] Arthur P Dempster, Nan M Laird, and Donald B Rubin. Maximum likelihood from incom- plete data via the em algorithm.Journal of the royal statis- tical society: series B (methodological), 39(1):1–22,

    1977

Show all 84 references
  1. [14]

    B.2 Real-world Experiments Dataset.To evaluate the performance of the proposed method, we consider the following datasets: 1)ETT[Zhouet al., 2021] is an electricity transformer temperature dataset collected from two separated counties in China, including four different dataset...

    2021

  2. [15]

    Bayotide: Bayesian online multivari- ate time series imputation with functional decomposition

    [Fanget al., 2023 ] Shikai Fang, Qingsong Wen, Shandian Zhe, and Liang Sun. Bayotide: Bayesian online multivari- ate time series imputation with functional decomposition. arXiv preprint arXiv:2308.14906,

    2023 arXiv

  3. [16]

    Gp-vae: Deep proba- bilistic time series imputation

    [Fortuinet al., 2020 ] Vincent Fortuin, Dmitry Baranchuk, Gunnar R¨atsch, and Stephan Mandt. Gp-vae: Deep proba- bilistic time series imputation. InInternational conference on artificial intelligence and statistics, pages 1651–1661. PMLR,

    2020

  4. [20]

    Hidden markov nonlinear ica: Unsupervised learning from nonstationary time series

    [H¨alv¨a and Hyvarinen, 2020] Hermanni H ¨alv¨a and Aapo Hyvarinen. Hidden markov nonlinear ica: Unsupervised learning from nonstationary time series. InConference on Uncertainty in Artificial Intelligence, pages 939–948. PMLR,

    2020

  5. [21]

    Latent processes identi- fication from multi-view time series.arXiv preprint arXiv:2305.08164,

    [Huanget al., 2023 ] Zenan Huang, Haobo Wang, Junbo Zhao, and Nenggan Zheng. Latent processes identi- fication from multi-view time series.arXiv preprint arXiv:2305.08164,

    2023 arXiv

  6. [22]

    Unsupervised feature extraction by time- contrastive learning and nonlinear ica.Advances in neural information processing systems, 29,

    [Hyvarinen and Morioka, 2016] Aapo Hyvarinen and Hi- roshi Morioka. Unsupervised feature extraction by time- contrastive learning and nonlinear ica.Advances in neural information processing systems, 29,

    2016

  7. [27]

    Independent compo- nent analysis: recent advances.Philosophical Transac- tions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 371(1984):20110534,

    [Hyv¨arinen, 2013] Aapo Hyv ¨arinen. Independent compo- nent analysis: recent advances.Philosophical Transac- tions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 371(1984):20110534,

    1984

  8. [29]

    Mimic-iii, a freely ac- cessible critical care database.Scientific data, 3(1):1–9,

    [Johnsonet al., 2016 ] Alistair EW Johnson, Tom J Pollard, Lu Shen, Li-wei H Lehman, Mengling Feng, Mohammad Ghassemi, Benjamin Moody, Peter Szolovits, Leo An- thony Celi, and Roger G Mark. Mimic-iii, a freely ac- cessible critical care database.Scientific data, 3(1):1–9,

    2016

  9. [30]

    Proba- bilistic imputation for time-series classification with miss- ing data

    [Kimet al., 2023 ] SeungHyun Kim, Hyunsu Kim, Eunggu Yun, Hwangrae Lee, Jaehun Lee, and Juho Lee. Proba- bilistic imputation for time-series classification with miss- ing data. InInternational Conference on Machine Learn- ing, pages 16654–16667. PMLR,

    2023

  10. [31]

    Partial disentanglement for do- main adaptation

    [Konget al., 2022 ] Lingjing Kong, Shaoan Xie, Weiran Yao, Yujia Zheng, Guangyi Chen, Petar Stojanov, Victor Akin- wande, and Kun Zhang. Partial disentanglement for do- main adaptation. InInternational conference on machine learning, pages 11455–11472. PMLR,

    2022

  11. [32]

    Identification of nonlinear latent hierarchical models.Advances in Neural Information Processing Systems, 36:2010–2032,

    [Konget al., 2023 ] Lingjing Kong, Biwei Huang, Feng Xie, Eric Xing, Yuejie Chi, and Kun Zhang. Identification of nonlinear latent hierarchical models.Advances in Neural Information Processing Systems, 36:2010–2032,

    2023

  12. [33]

    Partial disentanglement via mechanism sparsity.arXiv preprint arXiv:2207.07732,

    [Lachapelle and Lacoste-Julien, 2022] S´ebastien Lachapelle and Simon Lacoste-Julien. Partial disentanglement via mechanism sparsity.arXiv preprint arXiv:2207.07732,

    2022 arXiv

  13. [34]

    Synergies between disentanglement and sparsity: Gen- eralization and identifiability in multi-task learning

    [Lachapelleet al., 2023 ] S´ebastien Lachapelle, Tristan Deleu, Divyat Mahajan, Ioannis Mitliagkas, Yoshua Bengio, Simon Lacoste-Julien, and Quentin Bertrand. Synergies between disentanglement and sparsity: Gen- eralization and identifiability in multi-task learning. In Intern...

    2023

  14. [35]

    Modeling long-and short-term temporal patterns with deep neural networks

    [Laiet al., 2018 ] Guokun Lai, Wei-Cheng Chang, Yiming Yang, and Hanxiao Liu. Modeling long-and short-term temporal patterns with deep neural networks. InThe 41st international ACM SIGIR conference on research & devel- opment in information retrieval, pages 95–104,

    2018

  15. [36]

    Springer,

    [Lee and Lee, 1998] Te-Won Lee and Te-Won Lee.Indepen- dent component analysis. Springer,

    1998

  16. [38]

    When and how: Learning identifiable latent states for nonstationary time series forecasting.arXiv preprint arXiv:2402.12767,

    [Liet al., 2024b ] Zijian Li, Ruichu Cai, Zhenhui Yang, Haiqin Huang, Guangyi Chen, Yifan Shen, Zhengming Chen, Xiangchen Song, and Kun Zhang. When and how: Learning identifiable latent states for nonstationary time series forecasting.arXiv preprint arXiv:2402.12767,

    arXiv

  17. [39]

    On the identification of temporal causal representation with instantaneous dependence

    [Liet al., 2025 ] Zijian Li, Yifan Shen, Kaitao Zheng, Ruichu Cai, Xiangchen Song, Mingming Gong, Guangyi Chen, and Kun Zhang. On the identification of temporal causal representation with instantaneous dependence. InThe Thirteenth International Conference on Learning Repre- se...

    2025

  18. [40]

    Citris: Causal identifiability from temporal intervened se- quences

    [Lippeet al., 2022 ] Phillip Lippe, Sara Magliacane, Sindy L¨owe, Yuki M Asano, Taco Cohen, and Stratis Gavves. Citris: Causal identifiability from temporal intervened se- quences. InInternational Conference on Machine Learn- ing, pages 13557–13603. PMLR,

    2022

  19. [41]

    Naomi: Non-autoregressive mul- tiresolution sequence imputation.Advances in neural in- formation processing systems, 32,

    [Liuet al., 2019 ] Yukai Liu, Rose Yu, Stephan Zheng, Eric Zhan, and Yisong Yue. Naomi: Non-autoregressive mul- tiresolution sequence imputation.Advances in neural in- formation processing systems, 32,

    2019

  20. [42]

    Pristi: A condi- tional diffusion framework for spatiotemporal imputation

    [Liuet al., 2023a ] Mingzhe Liu, Han Huang, Hao Feng, Leilei Sun, Bowen Du, and Yanjie Fu. Pristi: A condi- tional diffusion framework for spatiotemporal imputation. In2023 IEEE 39th International Conference on Data En- gineering (ICDE), pages 1927–1939. IEEE,

    1927

  21. [43]

    itransformer: Inverted transformers are effective for time series forecasting.arXiv preprint arXiv:2310.06625,

    [Liuet al., 2023b ] Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. itransformer: Inverted transformers are effective for time series forecasting.arXiv preprint arXiv:2310.06625,

    arXiv

  22. [44]

    Challenging common assumptions in the unsupervised learning of disentangled representations

    [Locatelloet al., 2019 ] Francesco Locatello, Stefan Bauer, Mario Lucic, Gunnar Raetsch, Sylvain Gelly, Bernhard Sch¨olkopf, and Olivier Bachem. Challenging common assumptions in the unsupervised learning of disentangled representations. Ininternational conference on machine l...

    2019

  23. [45]

    Multivariate time series imputation with generative adversarial networks.Advances in neural information processing systems, 31,

    [Luoet al., 2018 ] Yonghong Luo, Xiangrui Cai, Ying Zhang, Jun Xu, et al. Multivariate time series imputation with generative adversarial networks.Advances in neural information processing systems, 31,

    2018

  24. [46]

    Object-centric architectures enable efficient causal representation learning.arXiv preprint arXiv:2310.19054,

    [Mansouriet al., 2023 ] Amin Mansouri, Jason Hartford, Yan Zhang, and Yoshua Bengio. Object-centric architectures enable efficient causal representation learning.arXiv preprint arXiv:2310.19054,

    2023 arXiv

  25. [47]

    Learning to reconstruct missing data from spa- tiotemporal graphs with sparse observations.Advances in Neural Information Processing Systems, 35:32069–32082,

    [Mariscaet al., 2022 ] Ivan Marisca, Andrea Cini, and Ce- sare Alippi. Learning to reconstruct missing data from spa- tiotemporal graphs with sparse observations.Advances in Neural Information Processing Systems, 35:32069–32082,

    2022

  26. [48]

    Generative semi-supervised learning for multivariate time series im- putation

    [Miaoet al., 2021 ] Xiaoye Miao, Yangyang Wu, Jun Wang, Yunjun Gao, Xudong Mao, and Jianwei Yin. Generative semi-supervised learning for multivariate time series im- putation. InProceedings of the AAAI conference on artifi- cial intelligence, volume 35, pages 8983–8991,

    2021

  27. [49]

    Graphical models for inference with missing data

    [Mohanet al., 2013 ] Karthika Mohan, Judea Pearl, and Jin Tian. Graphical models for inference with missing data. Advances in neural information processing systems, 26,

    2013

  28. [50]

    Uncertainty-aware variational- recurrent imputation network for clinical time series.IEEE Transactions on Cybernetics, 52(9):9684–9694,

    [Mulyadiet al., 2021 ] Ahmad Wisnu Mulyadi, Eunji Jun, and Heung-Il Suk. Uncertainty-aware variational- recurrent imputation network for clinical time series.IEEE Transactions on Cybernetics, 52(9):9684–9694,

    2021

  29. [51]

    Imputeformer: Low rankness-induced transformers for generalizable spatiotemporal imputation

    [Nieet al., 2023 ] Tong Nie, Guoyang Qin, Wei Ma, Yuewen Mei, and Jian Sun. Imputeformer: Low rankness-induced transformers for generalizable spatiotemporal imputation. arXiv: 2312.01728,

    2023 arXiv

  30. [52]

    Imputegan: Generative adversarial network for multivariate time series imputation.Entropy, 25(1):137,

    [Qin and Wang, 2023] Rui Qin and Yong Wang. Imputegan: Generative adversarial network for multivariate time series imputation.Entropy, 25(1):137,

    2023

  31. [53]

    Learning interpretable concepts: Unify- ing causal representation learning and foundation models

    [Rajendranet al., 2024 ] Goutham Rajendran, Simon Buch- holz, Bryon Aragam, Bernhard Sch ¨olkopf, and Pradeep Ravikumar. Learning interpretable concepts: Unify- ing causal representation learning and foundation models. arXiv preprint arXiv:2402.09236,

    2024 arXiv

  32. [54]

    Inference and missing data

    [Rubin, 1976] Donald B Rubin. Inference and missing data. Biometrika, 63(3):581–592,

    1976

  33. [57]

    [Simmonset al., 2010 ] Deborah Simmons, Angela Hairrell, Meaghan Edmonds, Sharon Vaughn, Ross Larsen, Victor Willson, William Rupley, and Glenda Byrns. A compari- son of multiple-strategy methods: Effects on fourth-grade students’ general and content-specific reading comprehen...

    2010

  34. [59]

    Probabilistic transformer for time series analy- sis.Advances in Neural Information Processing Systems, 34:23592–23608,

    [Tang and Matteson, 2021] Binh Tang and David S Matte- son. Probabilistic transformer for time series analy- sis.Advances in Neural Information Processing Systems, 34:23592–23608,

    2021

  35. [60]

    Csdi: Conditional score- based diffusion models for probabilistic time series impu- tation.Advances in Neural Information Processing Sys- tems, 34:24804–24816,

    [Tashiroet al., 2021 ] Yusuke Tashiro, Jiaming Song, Yang Song, and Stefano Ermon. Csdi: Conditional score- based diffusion models for probabilistic time series impu- tation.Advances in Neural Information Processing Sys- tems, 34:24804–24816,

    2021

  36. [61]

    mice: Multivariate imputation by chained equations in r.Journal of statistical software, 45:1–67,

    [Van Buuren and Groothuis-Oudshoorn, 2011] Stef Van Bu- uren and Karin Groothuis-Oudshoorn. mice: Multivariate imputation by chained equations in r.Journal of statistical software, 45:1–67,

    2011

  37. [63]

    Timesnet: Temporal 2d-variation modeling for general time series analysis.arXiv preprint arXiv:2210.02186,

    [Wuet al., 2022 ] Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. Timesnet: Temporal 2d-variation modeling for general time series analysis.arXiv preprint arXiv:2210.02186,

    2022 arXiv

  38. [64]

    Interpretable weather forecasting for worldwide stations with a unified deep model.Nature Ma- chine Intelligence, 5(6):602–611,

    [Wuet al., 2023 ] Haixu Wu, Hang Zhou, Mingsheng Long, and Jianmin Wang. Interpretable weather forecasting for worldwide stations with a unified deep model.Nature Ma- chine Intelligence, 5(6):602–611,

    2023

  39. [65]

    Multi-domain image generation and translation with identifiability guarantees

    [Xieet al., 2023 ] Shaoan Xie, Lingjing Kong, Mingming Gong, and Kun Zhang. Multi-domain image generation and translation with identifiability guarantees. InThe Eleventh International Conference on Learning Represen- tations,

    2023

  40. [66]

    Counterfac- tual generation with identifiability guarantees.Advances in Neural Information Processing Systems, 36,

    [Yanet al., 2024 ] Hanqi Yan, Lingjing Kong, Lin Gui, Yue- jie Chi, Eric Xing, Yulan He, and Kun Zhang. Counterfac- tual generation with identifiability guarantees.Advances in Neural Information Processing Systems, 36,

    2024

  41. [67]

    Learning temporally causal latent processes from general temporal data.arXiv preprint arXiv:2110.05428,

    [Yaoet al., 2021 ] Weiran Yao, Yuewen Sun, Alex Ho, Changyin Sun, and Kun Zhang. Learning temporally causal latent processes from general temporal data.arXiv preprint arXiv:2110.05428,

    2021 arXiv

  42. [68]

    Temporally disentangled representation learn- ing.Advances in Neural Information Processing Systems, 35:26492–26503,

    [Yaoet al., 2022 ] Weiran Yao, Guangyi Chen, and Kun Zhang. Temporally disentangled representation learn- ing.Advances in Neural Information Processing Systems, 35:26492–26503,

    2022

  43. [69]

    Multi-view causal representation learning with partial ob- servability.arXiv preprint arXiv:2311.04056,

    [Yaoet al., 2023 ] Dingling Yao, Danru Xu, S ´ebastien Lachapelle, Sara Magliacane, Perouz Taslakian, Georg Martius, Julius von K ¨ugelgen, and Francesco Locatello. Multi-view causal representation learning with partial ob- servability.arXiv preprint arXiv:2311.04056,

    2023 arXiv

  44. [70]

    Estimating missing data in temporal data streams using multi-directional recurrent neural net- works.IEEE Transactions on Biomedical Engineering, 66(5):1477–1490,

    [Yoonet al., 2018 ] Jinsung Yoon, William R Zame, and Mi- haela van der Schaar. Estimating missing data in temporal data streams using multi-directional recurrent neural net- works.IEEE Transactions on Biomedical Engineering, 66(5):1477–1490,

    2018

  45. [71]

    Kernel-based nonlinear independent component analysis

    [Zhang and Chan, 2007] Kun Zhang and Laiwan Chan. Kernel-based nonlinear independent component analysis. InInternational Conference on Independent Component Analysis and Signal Separation, pages 301–308. Springer,

    2007

  46. [73]

    Causal representation learning from multiple distributions: A general setting.arXiv preprint arXiv:2402.05052,

    [Zhanget al., 2024 ] Kun Zhang, Shaoan Xie, Ignavier Ng, and Yujia Zheng. Causal representation learning from multiple distributions: A general setting.arXiv preprint arXiv:2402.05052,

    2024 arXiv

  47. [74]

    On the identifiability of nonlinear ica: Sparsity and beyond.Advances in neural information processing systems, 35:16411–16422,

    [Zhenget al., 2022 ] Yujia Zheng, Ignavier Ng, and Kun Zhang. On the identifiability of nonlinear ica: Sparsity and beyond.Advances in neural information processing systems, 35:16411–16422,

    2022

  48. [75]

    Informer: Beyond efficient transformer for long sequence time-series forecasting

    [Zhouet al., 2021 ] Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wan- cai Zhang. Informer: Beyond efficient transformer for long sequence time-series forecasting. InProceedings of the AAAI conference on artificial intelligence, volume 35, pa...

    2021

  49. [76]

    This theory underpins estimation methods such as Maximum Likelihood [Dempsteret al., 1977 ], introduced in 1977, and Multiple Imputation [Rubin, 1978], developed by Rubin in

    A Related Works A.1 Time Series Imputation In the field of statistics, Rubin’s seminal work in 1976 laid a critical foundation for traditional missing data processing by establishing a framework of procedures and conditions designed to mitigate the adverse effects of missing d...

    1976

  50. [77]

    However, most traditional approaches to handling missing data rely on the assumptions of Missing Completely at Random (MCAR) or MAR

    These methods ensure convergence to consistent estimates under the Missing at Random (MAR) assumption. However, most traditional approaches to handling missing data rely on the assumptions of Missing Completely at Random (MCAR) or MAR. For instance, [Gibbet al., 2006 ] assumed...

    2006

  51. [81]

    Sincehc is invertible, thenct is component-wise identifiable

    That is, in each row ofJ A hc there is only one non-zero entry. Sincehc is invertible, thenct is component-wise identifiable. Theorem 2.(Identification of Latent States and Missing Causes under MNAR.)We follow the A1 in Theorem 1 and suppose that the observed data from missing...

    2022

  52. [82]

    Sincehc is invertible, thenct is component-wise identifiable

    That is, in each row ofJ B hc there is only one non-zero entry. Sincehc is invertible, thenct is component-wise identifiable. D More Details on Missing mechanism Since the missing mechanism in the imputation m-graph differs from the definition of standard m-graph [Mohanet al.,...

    2013

  53. [84]

    In a few real-world applications, certain violations of the assumption can be discernible to a certain degree, such as occlusion and visual persistence in videos

    G Limitation One limitation of our method is the invertible mixing process. In a few real-world applications, certain violations of the assumption can be discernible to a certain degree, such as occlusion and visual persistence in videos. For example, when an object is blocked...

    2024

  54. [1976]

    Multiple imputations in sample surveys-a phenomenological bayesian approach to nonresponse

    [Rubin, 1978] Donald B Rubin. Multiple imputations in sample surveys-a phenomenological bayesian approach to nonresponse. InProceedings of the survey research methods section of the American Statistical Association, volume 1, pages 20–34. American Statistical Association Alexa...

    1978

  55. [1977]

    Saits: Self-attention-based imputation for time series.Expert Systems with Applications, 219:119619,

    [Duet al., 2023 ] Wenjie Du, David Cˆot´e, and Yan Liu. Saits: Self-attention-based imputation for time series.Expert Systems with Applications, 219:119619,

    2023

  56. [1978]

    Toward causal representation learning.Proceedings of the IEEE, 109(5):612–634,

    [Sch¨olkopfet al., 2021 ] Bernhard Sch¨olkopf, Francesco Lo- catello, Stefan Bauer, Nan Rosemary Ke, Nal Kalch- brenner, Anirudh Goyal, and Yoshua Bengio. Toward causal representation learning.Proceedings of the IEEE, 109(5):612–634,

    2021

  57. [1994]

    Identifiability results for multimodal contrastive learning

    [Daunhaweret al., 2023 ] Imant Daunhawer, Alice Bizeul, Emanuele Palumbo, Alexander Marx, and Julia E V ogt. Identifiability results for multimodal contrastive learning. arXiv preprint arXiv:2303.09166,

    2023 arXiv

  58. [1998]

    Identifying semantic component for robust molecular property prediction.arXiv preprint arXiv:2311.04837,

    [Liet al., 2023 ] Zijian Li, Zunhong Xu, Ruichu Cai, Zhen- hui Yang, Yuguang Yan, Zhifeng Hao, Guangyi Chen, and Kun Zhang. Identifying semantic component for robust molecular property prediction.arXiv preprint arXiv:2311.04837,

    2023 arXiv

  59. [1999]

    Nonlinear ica using auxiliary vari- ables and generalized contrastive learning

    [Hyvarinenet al., 2019 ] Aapo Hyvarinen, Hiroaki Sasaki, and Richard Turner. Nonlinear ica using auxiliary vari- ables and generalized contrastive learning. InThe 22nd International Conference on Artificial Intelligence and Statistics, pages 859–868. PMLR,

    2019

  60. [2004]

    Diffusion-based time series impu- tation and forecasting with structured state space models

    [Alcaraz and Strodthoff, 2022] Juan Miguel Lopez Alcaraz and Nils Strodthoff. Diffusion-based time series impu- tation and forecasting with structured state space models. arXiv preprint arXiv:2208.09399,

    2022 arXiv

  61. [2006]

    Gender similari- ties and differences in the treatment, relapse, and recovery cycle.Evaluation review, 32(1):113–137,

    [Grellaet al., 2008 ] Christine E Grella, Christy K Scott, Mark A Foss, and Michael L Dennis. Gender similari- ties and differences in the treatment, relapse, and recovery cycle.Evaluation review, 32(1):113–137,

    2008

  62. [2007]

    Miss- ing value imputation in multivariate time series with end- to-end generative adversarial networks.Information Sci- ences, 551:67–82,

    [Zhanget al., 2021 ] Ying Zhang, Baohang Zhou, Xiangrui Cai, Wenya Guo, Xiaoke Ding, and Xiaojie Yuan. Miss- ing value imputation in multivariate time series with end- to-end generative adversarial networks.Information Sci- ences, 551:67–82,

    2021

  63. [2008]

    The incomplete rosetta stone problem: Iden- tifiability results for multi-view nonlinear ica

    [Greseleet al., 2020 ] Luigi Gresele, Paul K Rubenstein, Arash Mehrjou, Francesco Locatello, and Bernhard Sch¨olkopf. The incomplete rosetta stone problem: Iden- tifiability results for multi-view nonlinear ica. InUncer- tainty in Artificial Intelligence, pages 217–227. PMLR,

    2020

  64. [2010]

    Temporally disentangled rep- resentation learning under unknown nonstationarity.Ad- vances in Neural Information Processing Systems, 36,

    [Songet al., 2024 ] Xiangchen Song, Weiran Yao, Yewen Fan, Xinshuai Dong, Guangyi Chen, Juan Carlos Niebles, Eric Xing, and Kun Zhang. Temporally disentangled rep- resentation learning under unknown nonstationarity.Ad- vances in Neural Information Processing Systems, 36,

    2024

  65. [2011]

    Causal compo- nent analysis.Advances in Neural Information Processing Systems, 36,

    [Wendonget al., 2024 ] Liang Wendong, Armin Keki ´c, Julius von K ¨ugelgen, Simon Buchholz, Michel Besserve, Luigi Gresele, and Bernhard Sch ¨olkopf. Causal compo- nent analysis.Advances in Neural Information Processing Systems, 36,

    2024

  66. [2013]

    Pdformer: Propagation delay- aware dynamic long-range transformer for traffic flow pre- diction

    [Jianget al., 2023 ] Jiawei Jiang, Chengkai Han, Wayne Xin Zhao, and Jingyuan Wang. Pdformer: Propagation delay- aware dynamic long-range transformer for traffic flow pre- diction. InProceedings of the AAAI conference on artifi- cial intelligence, volume 37, pages 4365–4373,

    2023

  67. [2016]

    Nonlinear ica of temporally dependent sta- tionary sources

    [Hyvarinen and Morioka, 2017] Aapo Hyvarinen and Hi- roshi Morioka. Nonlinear ica of temporally dependent sta- tionary sources. InArtificial Intelligence and Statistics, pages 460–469. PMLR,

    2017

  68. [2017]

    Nonlinear independent component analysis: Existence and uniqueness results.Neural networks, 12(3):429–439,

    [Hyv¨arinen and Pajunen, 1999] Aapo Hyv¨arinen and Petteri Pajunen. Nonlinear independent component analysis: Existence and uniqueness results.Neural networks, 12(3):429–439,

    1999

  69. [2018]

    Recurrent neural networks for multivariate time series with missing values.Scientific reports, 8(1):6085,

    [Cheet al., 2018 ] Zhengping Che, Sanjay Purushotham, Kyunghyun Cho, David Sontag, and Yan Liu. Recurrent neural networks for multivariate time series with missing values.Scientific reports, 8(1):6085,

    2018

  70. [2019]

    Identifiability of latent- variable and structural-equation models: from linear to nonlinear.Annals of the Institute of Statistical Mathematics, 76(1):1–33,

    [Hyv¨arinenet al., 2024 ] Aapo Hyv ¨arinen, Ilyes Khe- makhem, and Ricardo Monti. Identifiability of latent- variable and structural-equation models: from linear to nonlinear.Annals of the Institute of Statistical Mathematics, 76(1):1–33,

    2024

  71. [2020]

    The hopeless- ness theory of depression: A prospective multi-wave test of the vulnerability-stress hypothesis.Cognitive Therapy and Research, 30:763–772,

    [Gibbet al., 2006 ] Brandon E Gibb, Christopher G Beevers, Margaret S Andover, and Kyle Holleran. The hopeless- ness theory of depression: A prospective multi-wave test of the vulnerability-stress hypothesis.Cognitive Therapy and Research, 30:763–772,

    2006

  72. [2021]

    Independent component anal- ysis, a new concept?Signal processing, 36(3):287–314,

    [Comon, 1994] Pierre Comon. Independent component anal- ysis, a new concept?Signal processing, 36(3):287–314,

    1994

  73. [2022]

    Identifiability of parameters in latent structure models with many observed variables

    [Allmanet al., 2009 ] Elizabeth S Allman, Catherine Matias, and John A Rhodes. Identifiability of parameters in latent structure models with many observed variables

    2009

  74. [2023]

    Caring: Learning temporal causal representation under non-invertible generation process

    [Chenet al., 2024 ] Guangyi Chen, Yifan Shen, Zhenhao Chen, Xiangchen Song, Yuewen Sun, Weiran Yao, Xiao Liu, and Kun Zhang. Caring: Learning temporal causal representation under non-invertible generation process. arXiv preprint arXiv:2401.14535,

    2024 arXiv

  75. [2024]

    Conditional information bottleneck approach for time se- ries imputation

    [Choi and Lee, 2023] MinGyu Choi and Changhee Lee. Conditional information bottleneck approach for time se- ries imputation. InThe Twelfth International Conference on Learning Representations,

    2023

  76. [2025]

    Learning disentangled representation for multi-modal time-series sensing signals

    [Caiet al., 2025b ] Ruichu Cai, Zhifan Jiang, Kaitao Zheng, Zijian Li, Weilin Chen, Xuexin Chen, Yifan Shen, Guangyi Chen, Zhifeng Hao, and Kun Zhang. Learning disentangled representation for multi-modal time-series sensing signals. InProceedings of the ACM on Web Con- ference...

    2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.