Pith. sign in

REVIEW 3 major objections 5 minor 65 references

Alternators With Noise Models

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 Alternator++ claims that learning the noise in observed and latent trajectories improves time-series density estimation, imputation, and forecasting beyond Mamba, ScoreGrad, and Dyffusion.

desk verdict The noise-matching loss in Alternator++ has no well-defined target, so the claimed mechanism is unsupported and the empirical wins are likely spurious. read the letter →

arxiv 2505.12544 v1 pith:INSEKHCP submitted 2025-05-18 cs.LG

Mohammad R. Rezaei , Adji Bousso Dieng This is my paper · ORCID
classification cs.LG
keywords Alternatorsnoisemodelstimeserieslatentdynamicsdensityestimationimputationforecastingdiffusion
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 tries to establish that a low-dimensional latent sequence model can capture complex, time-varying stochasticity by learning the noise that drives each observation and latent update. It adds two neural networks, one predicting observation noise and one predicting latent noise, and trains them with a noise-matching loss added to the Alternator objective. The reported result is that this model, Alternator++, gives lower MMD than Mamba, ScoreGrad, and the original Alternator on three of four density-estimation benchmarks, better imputation on most datasets, and lower MSE on sea-surface-temperature forecasting, while sampling far faster than diffusion alternatives. The practical payoff, if the results hold, is diffusion-level flexibility without a high-dimensional state or iterative sampling.

What carries the argument

The central machinery is the pair of trainable noise-prediction networks $\varepsilon^t_\psi$ and $\varepsilon^t_\nu$ added to the Alternator equations. An Alternator pairs an observed sequence $x_{1:T}$ with a low-dimensional latent trajectory $z_{0:T}$ and alternates between generating $x_t$ from $z_{t-1}$ and updating $z_t$ from $z_{t-1}$ and $x_t$; Alternator++ lets the Gaussian noise in both steps have learned means given by these networks. The noise-matching loss $L_\varepsilon$, balanced by $\gamma_t$, trains the networks to predict the Gaussian draws used in the sampling equations, which is the mechanism the paper credits for capturing time-varying stochasticity.

What would settle it

Calculate the minimizer of the noise-matching loss as written with $x_t$ fixed from the training data and $\varepsilon_x \sim \mathcal{N}(0,I)$: the conditional expectation of $\|\varepsilon_x - \varepsilon^t_\psi(z_{t-1})\|^2$ is minimized by the zero function, so a converged run should drive the observation noise network to zero unless the implemented objective differs from Eq. (10); reading the training code or checking the learned network outputs settles which case holds.

Watch

Extended reading notes

Core claim

At its core, the paper claims that the noise in an Alternator need not be a fixed zero-mean Gaussian: it can be learned by two small networks. The observation and latent updates are written as $$x_t = \sqrt{\beta_t} f_\$\theta$(z_{t-1}) + \sqrt{1-\beta_t-\$sigma_x^{2}$}\,\varepsilon^t_\psi(z_{t-1}) + \sigma_x \$varepsilon^{{\mu_x}}$_t$$ and $$z_t = \sqrt{\alpha_t} g_\varphi(x_t) + \sqrt{1-\alpha_t-\$sigma_z^{2}$}\,\varepsilon^t_\nu(z_{t-1}, x_t) + \sigma_z \$varepsilon^{{\mu_z}}$_t,$$ where $\varepsilon^t_\psi$ and $\varepsilon^t_\nu$ are the noise models. The added noise-matching loss trains these networks to reproduce the Gaussian draws used in those equations, and the paper argues this lets a low-dimensional latent Alternator capture complex stochasticity. Empirically, the paper reports lower MMD than Mamba, ScoreGrad, and the original Alternator on three of four density-estimation benchmarks, improved imputation on most datasets, and lower MSE on sea-surface-temperature forecasting, while avoiding the iterative sampling loop of diffusion models.

Load-bearing premise

The load-bearing premise is that the Gaussian noise samples drawn during training are exactly the noise that generated the observed and latent trajectories; the paper does not define how data would be turned into those specific noise samples, and the algorithm sometimes generates the observed values from the model rather than using the data.

Editorial extensions

If this is right

  • If the reported gains hold, low-dimensional latent states are enough for expressive stochastic time-series generation, reducing the need for high-dimensional hidden states in state-space models.
  • The noise-matching loss supplies a mechanism for modeling nonzero, state-dependent noise means, something the original Alternator's fixed zero-mean noise cannot represent.
  • At missing rates from 10% to 90%, the model would give imputation with lower mean squared error and higher correlation than Mamba and Alternator on most datasets tested.
  • For forecasting, the model would provide probabilistic forecasts with one-step alternating sampling, far cheaper than diffusion-based samplers: about 3x faster than Dyffusion and 50x faster than MCVD on the SST task.
  • The same generative process serves as an encoder for new sequences, so no separate inference network is needed for representation learning.

Reading between the lines

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

  • One extension beyond the paper is to define an explicit forward noising process so that $\varepsilon_x$ and $\varepsilon_z$ are genuinely the noises that produced the data; then the noise-matching loss becomes a denoising objective and the learned networks could be compared with diffusion score functions.
  • The recipe is portable: any latent dynamics model with Gaussian transition or observation noise could add a learned noise-prediction network and a matching loss as a cheap way to enrich its stochasticity.
  • Since the observation noise network conditions on $z_{t-1}$ only, a natural variation is to feed it the current observation $x_t$ as well, mirroring the latent noise network, and test whether imputation at high missing rates improves.
  • The paper notes the schedules $\beta_t$ and $\alpha_t$ need per-dataset tuning; learning them end-to-end rather than searching by hand would be the most direct next step.
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 Alternator++, an extension of the Alternator framework for time series modeling, which adds two trainable noise-prediction networks (one for observation noise, one for latent noise) and a noise-matching loss to the standard Alternator objective. The claimed mechanism is that this loss forces the noise networks to reproduce the Gaussian noise trajectories that generate the observed and latent sequences, thus enabling a low-dimensional latent Alternator to capture complex time-varying stochasticity while remaining computationally efficient. The authors report experiments on density estimation (MMD on Solar, Covid, Fred, NN5), imputation (MAE/MSE/CC at missing rates 10-90%), and sea-surface-temperature forecasting (CRPS/MSE), claiming improvements over Mamba, ScoreGrad, Dyffusion, and the original Alternator.

Significance. If the proposed noise-matching mechanism were valid, Alternator++ would be an interesting and useful contribution: it combines low-dimensional latent dynamics with learned noise models and shows promising computational advantages over diffusion-based baselines, and the empirical evaluation spans multiple tasks and datasets. However, the central training objective is degenerate as written: the targets of the noise-matching loss are independent standard Gaussian draws, so the loss cannot train the noise networks to approximate any data-dependent noise trajectories. This is a load-bearing flaw that invalidates the paper's core methodological claim, despite the plausibility of the overall research direction.

major comments (3)
  1. [Section 3.2, Eq. (10)] The noise-matching loss L_epsilon minimizes ||epsilon_z - eps_nu(z_{t-1}, x_t)||^2 + gamma_t ||epsilon_x - eps_psi(z_{t-1})||^2, where the text defines epsilon_x and epsilon_z as standard Gaussian variables sampled independently for each time step and batch element. Since these targets are independent of the networks' inputs, E[epsilon_x | z_{t-1}] = 0 and E[epsilon_z | z_{t-1}, x_t] = 0; the pointwise minimizers of the loss are therefore the zero functions. The loss cannot train the networks to approximate 'the noise trajectories that produce the observed and latent trajectories' because the paper never defines a forward process that determines those trajectories; no such well-defined targets exist. This degeneracy also holds under Algorithm 1, where x_t is a model sample independent of the freshly drawn epsilon_x. The central claimed mechanism of Alternator++ is thus unsupported.
  2. [Algorithm 1 vs. Section 3.2] Algorithm 1 samples the observation x_t from the generative process (x_t = mu_x_t + sigma_x eps_mu x_t), whereas Section 3.2 states that x_t is the t-th observation from the training data. Under the algorithm, the reconstruction term ||x_t - mu_x_t||^2 equals sigma_x^2 ||eps_mu x_t||^2, which is a random constant independent of the parameters; the Alternator loss would then provide no gradient signal for mu_x_t or mu_z_t. Under the text version, the reconstruction term is meaningful, but the noise-matching targets remain independent of the networks' inputs, so the noise-matching loss is still degenerate. The two descriptions of the training procedure are not equivalent, and the paper must specify which one is actually used.
  3. [Appendix B, Sea Surface Temperature Forecasting] The hyperparameter description for the SST forecasting experiment states that 'we train two Adversarial Diffusion Models (ADM)' with a U-Net backbone, base channels of 128, ResNet blocks, and 55.39M parameters; this does not correspond to the Alternator++ architecture described in Section 3.1, which uses the networks f_theta, g_phi, eps_psi, and eps_nu. This makes the reported SST forecasting results unreproducible from the methods as described and suggests that the experimental appendix does not accurately document the model used for those results.
minor comments (5)
  1. [Table 2 caption] The caption states that Alternator++ 'may underperform Mamba and Dyffusion in terms of CRPS,' but the reported CRPS values (0.212 for Alternator++ vs. 0.219 for Mamba and 0.224 for Dyffusion) indicate that Alternator++ has the lowest CRPS; this inconsistency should be clarified.
  2. [Section 3.2, notation] The symbols eps_x, eps_z, eps_mu x_t, eps_mu z_t, eps_psi, and eps_nu are used with overlapping notations; a table of notation would improve readability.
  3. [Section 5, related work] The claim that Alternator++ models state transitions in a 'non-Markovian way' is not supported, since the transition distribution depends on (z_{t-1}, x_t), which is a Markovian state in the joint state space.
  4. [Section 4.1, MMD evaluation] The MMD computation (kernel choice, bandwidth, number of sampled trajectories) is not described, which limits the reproducibility of the density estimation results.
  5. [Throughout] Several inline equations and formulas (e.g., Eq. (7)) contain garbled symbols or missing operators in the typeset version; the authors should carefully proofread the final manuscript.

Circularity Check

2 steps flagged · score 8.0 of 10

Noise-matching targets in Eq. (10) are independent Gaussian draws, making the optimal noise networks identically zero; Algorithm 1 compounds this by sampling the training observations from the model, so the claimed Alternator++ mechanism reduces to the original Alternator by construction.

  1. fitted input called prediction [Section 3.2, Eq. (10) and surrounding definitions; generative process Eqs. (8)-(9)]
    "The terms ε(b)x ∼ N(0, I Dx ) and ε(b)z ∼ N(0, I Dz ) are standard Gaussian noise variables sampled for each time step and batch element. ... Lε(θ,φ,ψ,ν) = 1/B Σ B b=1 Σ T t=1 (∥ε(b) z − εt ν(z(b) t−1, x(b) t)∥2 2 + γt· ∥ε(b) x − εt ψ(z(b) t−1)∥2 2)"

    By the paper's own definitions, ε_x and ε_z are freshly drawn N(0,I) variables, not the noises that generated the observed or latent trajectories. No forward process from data to these ε's is defined, and the text states that x_t is drawn from the training data. Since ε_x is independent of z_{t−1}, and ε_z is independent of (z_{t−1}, x_t), the conditional expectations of both targets are zero. The pointwise minimizers of the two squared terms are therefore ε_ψ≡0 and ε_ν≡0. The noise-matching loss provides no learnable signal, and its claimed 'prediction' of the data-generating noise is forced to zero by construction; the proposed mechanism cannot train the noise networks and Alternator++ collapses to Alternator with rescaled means.

  2. self definitional [Algorithm 1, lines 7-11]
    "Sample observation x (b) t =µ(b) x t +σxεµxt ... Compute lossL (θ,φ,ψ,ν) using (x1:T, z0:T,µz0:T ,µx1:T )"

    Here the 'observed' sequence used in the loss is produced by the model itself. Substituting x_t = µ_xt + σ_x ε_µxt into the Alternator observation term gives ||x_t − µ_xt||² = σ_x²||ε_µxt||², which is independent of θ, φ, ψ, and ν, so the observation reconstruction term contributes no gradient. The training target is thus defined as model output plus noise, making the objective tautological with respect to the observation term, and the data batch sampled at the start of Algorithm 1 never enters the loss. This contradicts the Section 3.2 statement that x_t is drawn from training data, and it makes the claimed fitting of noise trajectories self-referential and vacuous.

full rationale

The central contribution of Alternator++ is the noise-matching loss, and by the paper's own equations that loss targets independent standard Gaussian draws. The Bayes-optimal noise networks are identically zero, so the claimed mechanism is empty: with zero noise networks, the generative process is an Alternator with rescaled f/g means. Algorithm 1 makes the situation worse by defining the training observations as model samples, which makes the reconstruction term parameter-independent. These are not missing experiments or weak citations: the paper's own text defines the noise targets in a way that makes the 'prediction' equal to zero by construction, and defines the training observations as model-generated values in the algorithm. The empirical comparisons against Mamba, ScoreGrad, and Dyffusion are external benchmarks and would be meaningful as black-box model comparisons, and there is no load-bearing self-citation or imported uniqueness theorem; the circularity is internal to the derivation of the noise-modeling objective. The Limitations section only discusses schedule tuning and does not disclose this degeneracy, so the paper's asserted explanation of its reported gains is unsupported.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The model rests on the original Alternator objective (assumed valid) and on the paper-specific claim that independent Gaussian draws are the noise trajectories behind the data. The free parameters are the noise variances, schedules, and the mixing weight lambda, all chosen by hand or grid search per dataset. No new physical entities are introduced.

free parameters (5)
  • sigma_x = 0.3 (density), 0.15 (imputation), 0.2 (SST)
    Observation-space noise standard deviation, selected by grid search over {0.05,0.1,0.15,0.2,0.3} in Appendix B; the central model depends on it.
  • sigma_z = 0.15 (density/imputation), 0.1 (SST)
    Latent-space noise standard deviation, selected by grid search; affects all loss terms.
  • beta_t = linearly spaced schedule; exact values not reported
    Observation noise schedule that controls the balance between f_theta and epsilon_psi; the paper states it needs per-dataset tuning (Section 6).
  • alpha_t = linearly spaced for density/imputation; alpha_t=0.5 constant for SST
    Latent noise schedule, per-dataset tuned; admitted in Limitations section as domain-specific.
  • lambda = not reported
    Weight of the noise-matching loss; no value or search range given, but central to the objective (Eq. 10).
assumptions (3)
  • domain assumption The Alternator loss from Rezaei and Dieng (2024) is a valid training objective for the joint dynamics model.
    The paper inherits the Alternator training framework wholesale and does not re-derive it; the variant's soundness depends on the original framework.
  • ad hoc to paper The independent Gaussian variables epsilon_x and epsilon_z are the noise trajectories that produce the observed and latent data.
    This is the key premise of the noise-matching loss (Eq. 10). The data x_t is not generated by Eq. 8 in training (the text says x_t is from the data), so the noise that produced it is not defined; sampling epsilon_x independently makes it independent of the input to epsilon_psi.
  • ad hoc to paper A deterministic neural network can be trained to predict a target that is statistically independent of its input.
    The loss L_epsilon regresses epsilon_psi(z_{t-1}) onto epsilon_x ~ N(0,I). Since epsilon_x is independent of z_{t-1}, the optimal predictor is identically zero; the objective cannot drive the network to model noise trajectories.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Alternators With Noise Models." pith.science (2026). https://pith.science/paper/INSEKHCP

@misc{pith2026250512544,
  author       = {Pith},
  title        = {Pith review of: Alternators With Noise Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/INSEKHCP}},
  note         = {Machine review of arXiv:2505.12544}
}
read the original abstract

Alternators have recently been introduced as a framework for modeling time-dependent data. They often outperform other popular frameworks, such as state-space models and diffusion models, on challenging time-series tasks. This paper introduces a new Alternator model, called Alternator++, which enhances the flexibility of traditional Alternators by explicitly modeling the noise terms used to sample the latent and observed trajectories, drawing on the idea of noise models from the diffusion modeling literature. Alternator++ optimizes the sum of the Alternator loss and a noise-matching loss. The latter forces the noise trajectories generated by the two noise models to approximate the noise trajectories that produce the observed and latent trajectories. We demonstrate the effectiveness of Alternator++ in tasks such as density estimation, time series imputation, and forecasting, showing that it outperforms several strong baselines, including Mambas, ScoreGrad, and Dyffusion.

Figures

Figures reproduced from arXiv: 2505.12544 by the authors.

Figure 1
Figure 1. Comparing the distributions learned by various models against the ground truth distribution. Alternator++ captures multimodal distributions better than Alternator, Mamba, and ScoreGrad. sampling. We evaluate model robustness by varying the Missing At Random (MAR) rates from 10% to 90%. The results are summarized in [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Performance on missing data imputation across several datasets, evaluated in terms of MAE, MSE, and CC. Results are averaged over missing rates ranging from 10% to 90%. Alternator++ generally outperforms the baselines in terms of MSE and CC. However, for MAE, it faces challenges on the Covid dataset, where Alternator and Mamba perform better. to forecast SST using a daily dataset from 1982-2021, with data split into… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

65 extracted references · 20 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION fin.entry add.period write newline FUNCTION new.block output.state before.all = 'skip after.block 'output.state := if FUNCTION new.sentence output.state after.block = 'skip output.state before.all = 'skip after.sentence 'output.state := if if FUNCTION not #0 #1 if FUNCTION and 'skip pop #0 if FUNCTIO...

  2. [2]

    and Osendorfer, C

    Bayer, J. and Osendorfer, C. (2014). Learning stochastic recurrent networks. arXiv preprint arXiv:1411.7610

    arXiv 2014

  3. [3]

    R., Zhao, B., James, H., and Yu, R

    Cachay, S. R., Zhao, B., James, H., and Yu, R. (2023). Dyffusion: A dynamics-informed diffusion model for spatiotemporal forecasting. arXiv preprint arXiv:2306.01984

    arXiv 2023

  4. [4]

    Castrejon, L., Ballas, N., and Courville, A. (2019). Improved conditional vrnns for video prediction. In Proceedings of the IEEE/CVF international conference on computer vision , pages 7608--7617

    work page 2019

  5. [5]

    Chen, R., Duvenaud, D., and Rubanova, Y. (2019). Latent odes for irregularly-sampled time series. Advances in Neural Information Processing Systems , 32:3

    work page 2019

  6. [6]

    T., Rubanova, Y., Bettencourt, J., and Duvenaud, D

    Chen, R. T., Rubanova, Y., Bettencourt, J., and Duvenaud, D. K. (2018). Neural ordinary differential equations. Advances in neural information processing systems , 31

    2018

  7. [7]

    Chen, T. (2023). On the importance of noise scheduling for diffusion models. arXiv preprint arXiv:2301.10972

    arXiv 2023

  8. [8]

    C., and Bengio, Y

    Chung, J., Kastner, K., Dinh, L., Goel, K., Courville, A. C., and Bengio, Y. (2015). A recurrent latent variable model for sequential data. Advances in neural information processing systems , 28

    2015

Show all 65 references
  1. [9]

    S., Benidis, K., Bohlke-Schneider, M., Kurle, R., Stella, L., Hasson, H., Gallinari, P., and Januschowski, T

    de B \'e zenac, E., Rangapuram, S. S., Benidis, K., Bohlke-Schneider, M., Kurle, R., Stella, L., Hasson, H., Gallinari, P., and Januschowski, T. (2020). Normalizing kalman filters for multivariate time series analysis. Advances in Neural Information Processing Systems , 33:2995--3007

    2020

  2. [10]

    and Nichol, A

    Dhariwal, P. and Nichol, A. (2021). Diffusion models beat gans on image synthesis. Advances in neural information processing systems , 34:8780--8794

    2021

  3. [11]

    Doerr, A., Daniel, C., Schiegg, M., Duy, N.-T., Schaal, S., Toussaint, M., and Sebastian, T. (2018). Probabilistic recurrent state-space models. In International conference on machine learning , pages 1280--1289. PMLR

    2018

  4. [12]

    Dong, E., Du, H., and Gardner, L. (2020). An interactive web-based dashboard to track covid-19 in real time. The Lancet infectious diseases , 20(5):533--534

    2020

  5. [13]

    and Van Amersfoort, J

    Fabius, O. and Van Amersfoort, J. R. (2014). Variational recurrent auto-encoders. arXiv preprint arXiv:1412.6581

    2014 arXiv

  6. [14]

    Fortunato, M., Blundell, C., and Vinyals, O. (2017). Bayesian recurrent neural networks. arXiv preprint arXiv:1704.02798

    2017 arXiv

  7. [15]

    K., Paquet, U., and Winther, O

    Fraccaro, M., S nderby, S. K., Paquet, U., and Winther, O. (2016). Sequential neural models with stochastic layers. Advances in neural information processing systems , 29

    2016

  8. [16]

    and Dieng, A

    Friedman, D. and Dieng, A. B. (2023). The Vendi Score: A Diversity Evaluation Metric for Machine Learning . Transactions on Machine Learning Research

    2023

  9. [17]

    and Ghahramani, Z

    Gal, Y. and Ghahramani, Z. (2016). Dropout as a bayesian approximation: Representing model uncertainty in deep learning. In international conference on machine learning , pages 1050--1059. PMLR

    2016

  10. [18]

    and Katzfuss, M

    Gneiting, T. and Katzfuss, M. (2014). Probabilistic forecasting. Annual Review of Statistics and Its Application , 1:125--151

    2014

  11. [19]

    I., Hyndman, R

    Godahewa, R., Bergmeir, C., Webb, G. I., Hyndman, R. J., and Montero-Manso, P. (2021). Monash time series forecasting archive. arXiv preprint arXiv:2105.06643

    2021 arXiv

  12. [20]

    Gong, S., Li, M., Feng, J., Wu, Z., and Kong, L. (2022). Diffuseq: Sequence to sequence text generation with diffusion models. arXiv preprint arXiv:2210.08933

    2022 arXiv

  13. [21]

    and Dao, T

    Gu, A. and Dao, T. (2023). Mamba: Linear-time sequence modeling with selective state spaces. arXiv preprint arXiv:2312.00752

    2023 arXiv

  14. [22]

    Gu, A., Goel, K., and Ré, C. (2022). Efficiently modeling long sequences with structured state spaces

    2022

  15. [23]

    Haghbin, M., Sharafati, A., Motta, D., Al-Ansari, N., and Noghani, M. H. M. (2021). Applications of soft computing models for predicting sea surface temperature: a comprehensive review and assessment. Progress in earth and planetary science , 8:1--19

    2021

  16. [24]

    P., Poole, B., Norouzi, M., Fleet, D

    Ho, J., Chan, W., Saharia, C., Whang, J., Gao, R., Gritsenko, A., Kingma, D. P., Poole, B., Norouzi, M., Fleet, D. J., et al. (2022). Imagen video: High definition video generation with diffusion models. arXiv preprint arXiv:2210.02303

    2022 arXiv

  17. [25]

    Ho, J., Jain, A., and Abbeel, P. (2020). Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems , 33:6840--6851

    2020

  18. [26]

    Karl, M., Soelch, M., Bayer, J., and Van der Smagt, P. (2016). Deep variational bayes filters: Unsupervised learning of state space models from raw data. arXiv preprint arXiv:1605.06432

    2016 arXiv

  19. [27]

    Karras, T., Aittala, M., Aila, T., and Laine, S. (2022). Elucidating the design space of diffusion-based generative models. Advances in neural information processing systems , 35:26565--26577

    2022

  20. [28]

    Kong, Z., Ping, W., Huang, J., Zhao, K., and Catanzaro, B. (2020). Diffwave: A versatile diffusion model for audio synthesis. arXiv preprint arXiv:2009.09761

    2020 arXiv

  21. [29]

    G., Shalit, U., and Sontag, D

    Krishnan, R. G., Shalit, U., and Sontag, D. (2015). Deep kalman filters. arXiv preprint arXiv:1511.05121

    2015 arXiv

  22. [30]

    Lai, G., Chang, W.-C., Yang, Y., and Liu, H. (2017). Modeling long-and short-term temporal patterns with deep neural networks. corr abs/1703.07015 (2017). arXiv preprint arXiv:1703.07015

    2017 arXiv

  23. [31]

    Lee, S., Lee, K., and Park, T. (2024). Ant: Adaptive noise schedule for time series diffusion models. arXiv preprint arXiv:2410.14488

    2024 arXiv

  24. [32]

    B., and Beeson, R

    Li, A., Ding, Z., Dieng, A. B., and Beeson, R. (2024). Constraint-aware diffusion models for trajectory optimization. arXiv preprint arXiv:2406.00990

    2024 arXiv

  25. [33]

    Lin, S., Liu, B., Li, J., and Yang, X. (2024). Common diffusion noise schedules and sample steps are flawed. In Proceedings of the IEEE/CVF winter conference on applications of computer vision , pages 5404--5411

    2024

  26. [34]

    Liu, X., Xiao, T., Si, S., Cao, Q., Kumar, S., and Hsieh, C.-J. (2019). Neural sde: Stabilizing neural ode networks with stochastic noise

    2019

  27. [35]

    Lu, C., Zhou, Y., Bao, F., Chen, J., Li, C., and Zhu, J. (2022). Dpm-solver: A fast ode solver for diffusion probabilistic model sampling in around 10 steps. arXiv preprint arXiv:2206.00927

    2022 arXiv

  28. [36]

    J., and Romero, J

    Martinez, J., Black, M. J., and Romero, J. (2017). On human motion prediction using recurrent neural networks. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 2891--2900

    2017

  29. [37]

    Matheson, J. E. and Winkler, R. L. (1976). Scoring rules for continuous probability distributions. Management science , 22(10):1087--1096

    1976

  30. [38]

    McCracken, M. W. and Ng, S. (2016). Fred-md: A monthly database for macroeconomic research. Journal of Business & Economic Statistics , 34(4):574--589

    2016

  31. [39]

    Nichol, A. Q. and Dhariwal, P. (2021). Improved denoising diffusion probabilistic models. In International Conference on Machine Learning , pages 8162--8171. PMLR

    2021

  32. [40]

    Pathak, J., Subramanian, S., Harrington, P., Raja, S., Chattopadhyay, A., Mardani, M., Kurth, T., Hall, D., Li, Z., Azizzadenesheli, K., et al. (2022). Fourcastnet: A global data-driven high-resolution weather model using adaptive fourier neural operators. arXiv preprint arXiv...

    2022 arXiv

  33. [41]

    S., Seeger, M

    Rangapuram, S. S., Seeger, M. W., Gasthaus, J., Stella, L., Wang, Y., and Januschowski, T. (2018). Deep state space models for time series forecasting. Advances in neural information processing systems , 31

    2018

  34. [42]

    Rasul, K., Seward, C., Schuster, I., and Vollgraf, R. (2021). Autoregressive denoising diffusion models for multivariate probabilistic time series forecasting. In International Conference on Machine Learning , pages 8857--8868. PMLR

    2021

  35. [43]

    R., Arai, K., Frank, L

    Rezaei, M. R., Arai, K., Frank, L. M., Eden, U. T., and Yousefi, A. (2021). Real-time point process filter for multidimensional decoding problems using mixture models. Journal of neuroscience methods , 348:109006

    2021

  36. [44]

    Rezaei, M. R. and Dieng, A. B. (2024). Alternators for sequence modeling. arXiv preprint arXiv:2405.11848

    2024 arXiv

  37. [45]

    Rezaei, M. R. and Dieng, A. B. (2025). The alpha -alternator: Dynamic adaptation to varying noise levels in sequences using the vendi score for improved robustness and performance. arXiv preprint arXiv:2502.04593

    2025 arXiv

  38. [46]

    R., Hadjinicolaou, A

    Rezaei, M. R., Hadjinicolaou, A. E., Cash, S. S., Eden, U. T., and Yousefi, A. (2022). Direct discriminative decoder models for analysis of high-dimensional dynamical neural data. Neural Computation , 34(5):1100--1135

    2022

  39. [47]

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., and Ommer, B. (2022). High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 10684--10695

    2022

  40. [48]

    T., and Duvenaud, D

    Rubanova, Y., Chen, R. T., and Duvenaud, D. K. (2019). Latent ordinary differential equations for irregularly-sampled time series. Advances in neural information processing systems , 32

    2019

  41. [49]

    and Ho, J

    Salimans, T. and Ho, J. (2022). Progressive distillation for fast sampling of diffusion models. arXiv preprint arXiv:2202.00512

    2022 arXiv

  42. [50]

    and Kwok, J

    Shen, L. and Kwok, J. (2023). Non-autoregressive conditional diffusion models for time series prediction. In International Conference on Machine Learning , pages 31016--31029. PMLR

    2023

  43. [51]

    Shi, Z. (2024). Mambastock: Selective state space model for stock prediction. arXiv preprint arXiv:2402.18959

    2024 arXiv

  44. [52]

    Smith, J. T. H., Warrington, A., and Linderman, S. W. (2023). Simplified state space layers for sequence modeling

    2023

  45. [53]

    Song, J., Meng, C., and Ermon, S. (2020). Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502

    2020 arXiv

  46. [54]

    B., Bontempi, G., Atiya, A

    Taieb, S. B., Bontempi, G., Atiya, A. F., and Sorjamaa, A. (2012). A review and comparison of strategies for multi-step ahead time series forecasting based on the nn5 forecasting competition. Expert systems with applications , 39(8):7067--7083

    2012

  47. [55]

    Tashiro, Y., Song, J., Song, Y., and Ermon, S. (2021). Csdi: Conditional score-based diffusion models for probabilistic time series imputation. Advances in Neural Information Processing Systems , 34:24804--24816

    2021

  48. [56]

    Vahdat, A., Kreis, K., and Kautz, J. (2021). Score-based generative modeling in latent space. Advances in Neural Information Processing Systems , 34:11287--11302

    2021

  49. [57]

    Voleti, V., Jolicoeur-Martineau, A., and Pal, C. (2022). Mcvd-masked conditional video diffusion for prediction, generation, and interpolation. Advances in Neural Information Processing Systems , 35:23371--23385

    2022

  50. [58]

    Wang, Z., Kong, F., Feng, S., Wang, M., Yang, X., Zhao, H., Wang, D., and Zhang, Y. (2025). Is mamba effective for time series forecasting? Neurocomputing , 619:129178

    2025

  51. [59]

    Xu, B., Lu, Y., Inoue, Y., Lee, N., Fu, T., and Chen, J. (2024). Protein-mamba: Biological mamba models for protein function prediction. arXiv preprint arXiv:2409.14617

    2024 arXiv

  52. [60]

    Yan, T., Zhang, H., Zhou, T., Zhan, Y., and Xia, Y. (2021). Scoregrad: Multivariate probabilistic time series forecasting with continuous energy-based generative models. arXiv preprint arXiv:2106.10121

    2021 arXiv

  53. [61]

    Ye, Z., Chen, Z., Li, T., Huang, Z., Luo, W., and Qi, G.-J. (2024). Schedule on the fly: Diffusion time prediction for faster and better image generation. arXiv preprint arXiv:2412.01243

    2024 arXiv

  54. [62]

    Zhu, L., Liao, B., Zhang, Q., Wang, X., Liu, W., and Wang, X. (2024). Vision mamba: Efficient visual representation learning with bidirectional state space model. arXiv preprint arXiv:2401.09417

    2024 arXiv

  55. [63]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  56. [64]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  57. [65]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

Pith tools

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