Pith. sign in

REVIEW 4 major objections 4 minor 70 references

Generative Unordered Flow for Set-Structured Data Generation

T0 review · 4 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

Pith's one-line read Flow model that treats point sets as unordered beats baselines

desk verdict Sound core idea and strong empirical results, but the decoding step lacks a correctness guarantee and the 'first' claim is overstated. read the letter →

arxiv 2501.17770 v2 pith:44EDHEOK submitted 2025-01-29 cs.LG

classification cs.LG
keywords flowmatchingset-structureddatapermutationinvarianceGaussianmixturerepresentationpointprocessesLangevindynamicsfunction-valuedgenerativemodelssetgeneration
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

This paper attempts to establish that set-structured data—point sets with no inherent order—can be generated by a flow-based generative model without ever imposing an ordering on the points. The central move is to encode each set as a smooth function, namely a Gaussian mixture centered at the points, to train a flow-matching model on the probability measure of such functions, and then to decode a generated function back into a point set by particle search. The paper claims this is the first flow-based generative model that is permutation-invariant to unordered data. It reports lower discrepancy scores than point-process, energy-based, normalizing-flow, and diffusion baselines on three real-world spatiotemporal point-set datasets.

What carries the argument

The load-bearing object is the Gaussian-mixture representation $f_{X,\sigma(\epsilon)}$, which converts each set into a density-like element of $L^2(\mathbb{R}^{D_X})$. This permits the authors to transplant Euclidean flow matching to a Hilbert space: a neural operator parameterizes a function-valued vector field $u_{\theta,t}$, the conditional flow is the linear interpolation $\varphi_t(h)=(1-(1-\zeta)t)h+t h_{\mathrm{cond}}$, and training minimizes the $L^2$ squared error between the predicted and the closed-form conditional vector fields. The same representation powers decoding: for small $\epsilon$, the points of $X$ sit at the local maxima of the mixture, so Langevin warm-up followed by gradient ascent and single-pass clustering recovers them.

What would settle it

Build a synthetic Gaussian mixture with two modes separated by roughly $\epsilon$ plus a low, broad shoulder of comparable total mass, generate the function exactly from the training representation, and run the inverse transform. If the decoded set merges the two nearby points into one or reports the shoulder as a point, the local-maximum-plus-filtering recovery premise fails in the method's own operating regime.

Watch

Extended reading notes

Core claim

The central claim is that any finite unordered point set $X$ can be represented as the Gaussian mixture $f_{X,\sigma(\epsilon)} = \frac{1}{N}\sum_{i=1}^{N} G(x_i, \sigma_i(\epsilon)^2 I)$ with adaptive variances $\sigma_i(\epsilon)=\epsilon\ln(1+\min_{j\neq i}\|x_i-x_j\|_2)$, and that this representation converges weakly to the exact Dirac-delta representation $f_X$ as $\epsilon\to 0$, with a Wasserstein convergence rate of $O(\epsilon\ln\rho\sqrt{D_P})$. Because such mixtures are square-integrable, their probability measure is supported on the Hilbert space $L^2(\mathbb{R}^{D_X})$, where function-valued flow matching can be run. The inverse map is a particle-filtering-style procedure: initialize many particles, warm them up with Langevin dynamics so they concentrate near high density, move them by gradient ascent to local maxima, and merge and filter nearby particles to obtain the point set. The paper argues that this pipeline is permutation-invariant by construction and that the experiments support its claim of improved generation quality relative to prior baselines.

Load-bearing premise

The method assumes that a generated function is close enough to a well-separated Gaussian mixture that every true point is a local maximum and every spurious peak is low and small enough for clustering to discard; the paper does not bound how often this fails.

Editorial extensions

If this is right

  • A point-set generator invariant to the order of its inputs can be trained with the same regression objective as Euclidean flow matching, so existing flow-matching machinery transfers to unordered data.
  • The encode-generate-decode template applies any time the target is an unordered collection rather than an ordered vector, including conditional generation and interpolation between point sets.
  • Because the probability measure lives on $L^2(\mathbb{R}^{D_X})$, the middle stage can reuse function-space generative tools and Gaussian-measure priors instead of requiring bespoke discrete diffusion over points.
  • On the three real datasets tested, the claimed advantage is consistent on both a set-size metric and a point-location metric, which is the practical payoff the paper advertises.

Reading between the lines

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

  • Going beyond the paper, the set-specific burden is entirely in the encoder (set into mixture) and decoder (mixture into set), so the middle stage is generic function-space generative modeling; replacing flow matching with another $L^2$-supported generative model should work unchanged.
  • The decoding step is the least-supported part of the pipeline, since the convergence proof covers the representation but not the failure rate of local-maximum recovery; a natural testable extension is to anneal $\epsilon$ during Langevin warm-up or decode at multiple resolutions before merging.
  • A conjecture implicit in the setup is that the advantage over baselines grows as cardinality varies widely, because the representation couples point count to mixture weights; datasets with near-constant set sizes may show smaller gains.
  • The synthetic experiments use smooth intensities; anisotropic or strongly clustered point arrangements would stress the adaptive-variance assumption more than the tested Poisson and Hawkes cases.
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.

Referee Report

4 major / 4 minor

Summary. The paper proposes 'unordered flow,' a generative model for set-structured data. Sets are represented as normalized Gaussian mixtures with adaptive variances (Eq. 7), the distribution of such functions is modeled by a function-space flow-matching model trained with an L2 loss (Eq. 14), and a sampled function is decoded into a point set by initializing particles, warming them up with Langevin dynamics (Eq. 16), performing gradient ascent (Eq. 15), and clustering/deduplicating the converged particles. The paper proves weak convergence of the representation to a delta sum and L2 membership of the mixture space, and reports experiments on two synthetic point processes and three real-world datasets, with ablation studies for the main components.

Significance. If the decoding step were reliable, the proposed approach would be a useful contribution: it avoids the need for point ordering, handles variable set sizes, and provides a clean function-space formulation whose convergence and L2 support are proven in the appendix. The representation idea is simple and the theoretical statements in Propositions 3.1 and 3.2 are essentially correct. The paper is self-contained, includes proofs, and evaluates each module through ablations. However, the central inverse transform is only heuristically justified, and the paper's own caveats in Section 3.3 admit that noisy peaks and non-mixture-like generated functions can break decoding. In addition, the claimed novelty ('first flow-based generative model that is permutation-invariant') is contradicted by the paper's own baseline, Biloš (2021), which is described as a permutation-invariant normalizing flow. The empirical claim of 'significantly outperforming' baselines is not backed by error bars or significance tests.

major comments (4)
  1. [Section 3.3, Eq. (16)] The inverse transform is undefined on non-positive functions. Langevin warm-up in Eq. (16) uses the gradient of log bf_{*,σ(ε)}, but nothing in the function-valued flow matching framework guarantees that a generated function is positive. The initial Gaussian measure η0 produces draws that can be negative, and the conditional interpolation in Eq. (12), followed by a model trained with the L2 loss in Eq. (14), imposes no sign constraint on the output. The paper acknowledges in Section 3.3 that imperfectly trained models can produce non-mixture-like functions, but it does not address this domain problem. Since decoding is the step that turns a sampled function into the final point set, this is a load-bearing gap.
  2. [Section 3.3 and Proposition 3.1] No finite-ε recovery guarantee is provided. Proposition 3.1 states that f_{X,σ(ε)} converges weakly to the delta representation as ε → 0, with a Wasserstein rate, but this does not imply that the local maxima of f_{X,σ(ε)} coincide with X for the fixed ε used in training, nor does it bound the error of the decoded point set. The adaptive variance in Eq. (7) does mitigate mode merging for small ε (one has σ_i < ε d_i, so for ε < 1/2 neighboring oracle components remain separated), but the paper gives no statement about spurious modes, missed points, or the failure rate of the decoder when the generated function is not an exact mixture. The ablation in Table 2 shows sensitivity to the decoding components but does not quantify the failure rate.
  3. [Abstract and Section 4] The claim that unordered flow is 'the first flow-based generative model that is permutation-invariant to unordered data' is contradicted by the paper's own related work and baseline. Section 4 describes Biloš (2021) as a normalizing flow that 'learned a permutation invariant density distribution,' and Table 1 includes it as 'Permutation-invariant Normalizing Flow.' If the intended claim is limited to flow matching or to continuous flows in function space, that should be stated precisely; otherwise the novelty claim is not accurate.
  4. [Section 5.3, Table 1] The abstract and Section 5.3 claim that the model 'significantly outperforms' previous baselines, but the reported evidence does not support the word 'significantly' in a statistical sense. Table 1 shows single average values over 10 runs, with no error bars, confidence intervals, or significance tests. In addition, two baseline results are copied from Lüdke et al. (2024) while the other two come from the authors' own implementation, and no protocol is given for tuning or for ensuring comparable training conditions. The relative improvements may be real, but the significance claim is not established by the reported numbers.
minor comments (4)
  1. [Section 3.3, Eq. (16)] The noise term in the Langevin update is written as z_i^{(s)} ∼ G(1, I) and then described as 'a standard Gaussian noise.' A standard Gaussian has mean zero, so it should be G(0, I), not G(1, I).
  2. [Section 3.2, Eq. (11)] For η0 = G(0, Γ) to be a valid Gaussian measure on the infinite-dimensional Hilbert space L2(R^{D_X}), the covariance operator Γ must be trace-class. The paper only says Γ is 'well-defined (e.g., symmetric)'; please specify the trace-class condition or cite the relevant existence theorem.
  3. [Proposition 3.1 and Appendix A] The notation for the data dimension is inconsistent: Proposition 3.1 and Appendix A use D_P while the main text uses D_X. Please unify the notation.
  4. [Section 5, evaluation metrics] The metric name 'S-WStein' is used without defining the abbreviation. Please spell out what the metric measures and how the Wasserstein distance is estimated, and clarify whether the reported numbers are means, medians, or best runs.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the encoding-flow-decoding pipeline is an explicit construction evaluated against external baselines, with no self-citation or fitted-prediction loop.

full rationale

The paper's derivation chain is self-contained and does not reduce to its own inputs. The set-to-function map (Eq. 6-7) explicitly constructs a Gaussian mixture representation from the point set X, the function-valued flow matching objective (Eq. 14) is the standard conditional flow matching loss adapted to L2, and the inverse transform (Sec. 3.3) is a fixed, untrained decoding procedure consisting of Langevin warm-up, gradient ascent, and clustering. No parameter is fitted to the evaluation metrics, no 'prediction' is a renamed fit, and the central claims are not justified by self-citation: the reference list contains no papers by the current authors. The only caveat is that decoding generated functions into point sets relies on the heuristic assumption that modes of the generated function correspond to true points; the paper itself acknowledges this sensitivity and addresses it with ablations and a case study. That is a correctness and robustness concern, not a circularity, because the representation, training loss, and decoding are independently specified and the empirical evaluation compares decoded sets against external baselines. Accordingly, no specific circular step can be quoted, and the score is 0.

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

The central claim rests on several hyperparameters that are not reported, the ad hoc variance-scaling formula, and the assumption that the trained generative functions remain decodable by local-maxima search. No new physical entities are introduced.

free parameters (5)
  • epsilon (Gaussian bandwidth) = not reported
    Controls the width of the Gaussian bumps in the mixture representation (Eq. 6-7). The paper gives no value, and the ablation shows the method is sensitive to this choice.
  • sigma_i scaling (log function) = not reported
    The log-scaling function in Eq. (7) is an ad hoc choice, chosen for the convergence proof; no justification for the logarithmic form.
  • Inverse transform hyperparameters (Slgvin, Sgrad, alpha, beta, M) = not reported
    Number of Langevin steps, gradient steps, step sizes, and particle count are not specified in the paper.
  • Covariance Gamma of initial Gaussian measure = not specified
    The initial noise measure eta0 = G(0, Gamma) in the function space requires choosing the covariance operator Gamma, which is not described.
  • Neural network architecture and training hyperparameters = not reported
    No details on the vector field model, optimizer, or training loop.
assumptions (5)
  • standard math Gaussian densities converge weakly to Dirac deltas as variance tends to zero
    Used in Appendix A.1 to prove the mixture representation's weak convergence.
  • standard math The product of two Gaussian densities is a Gaussian with a combined variance term
    Used in Appendix B to show square-integrability of the mixture representation.
  • domain assumption Sets are finite and generated by an inhomogeneous Poisson process with regular intensity
    Stated in Proposition 3.1 and its proof to justify N < infinity almost surely.
  • domain assumption Functional flow matching theory (Kerrigan et al., 2024; Lim et al., 2023) applies to the L2 Hilbert space with a Gaussian initial measure
    The paper relies on these prior works for the correctness of function-valued flow matching; the extension is not proven in this paper.
  • ad hoc to paper The trained model produces functions sufficiently close to Gaussian mixtures that gradient ascent on local maxima recovers the point set
    The inverse transform's correctness is assumed; the paper only offers empirical evidence.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Generative Unordered Flow for Set-Structured Data Generation." pith.science (2026). https://pith.science/paper/44EDHEOK

@misc{pith2026250117770,
  author       = {Pith},
  title        = {Pith review of: Generative Unordered Flow for Set-Structured Data Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/44EDHEOK}},
  note         = {Machine review of arXiv:2501.17770}
}
read the original abstract

Flow-based generative models have demonstrated promising performance across a broad spectrum of data modalities (e.g., image and text). However, there are few works exploring their extension to unordered data (e.g., spatial point set), which is not trivial because previous models are mostly designed for vector data that are naturally ordered. In this paper, we present unordered flow, a type of flow-based generative model for set-structured data generation. Specifically, we convert unordered data into an appropriate function representation, and learn the probability measure of such representations through function-valued flow matching. For the inverse map from a function representation to unordered data, we propose a method similar to particle filtering, with Langevin dynamics to first warm-up the initial particles and gradient-based search to update them until convergence. We have conducted extensive experiments on multiple real-world datasets, showing that our unordered flow model is very effective in generating set-structured data and significantly outperforms previous baselines.

Figures

Figures reproduced from arXiv: 2501.17770 by the authors.

Figure 1
Figure 1. Our unordered flow model uθ,t applied to an inhomogeneous Poisson process. The left two subfigures illustrates the Poisson intensity function, with a mixture representation fb ⋆,σ(ϵ) generated from the model. The right two respectively depict the warm-up of uniformly initialized particles Y(1) and their convergence to the point set Xb . of methods in fact differ from the conventional diffusion models (which are base… view at source ↗
Figure 2
Figure 2. Results from our unordered flow model uθ,t on the Hawkes process. The left two subfigures illustrate a point set X in the training set, with a mixture representation fb ⋆,σ(ϵ) sampled from the model. The right two show the empirical distributions of particles, including their initial Y(1), warmed-up Y(Slgvin) , and converged versions Yb . Method Earthquakes COVID-19 Citibike S-WStein D-MMD S-WStein D-MMD S-WStein D-… view at source ↗
Figure 3
Figure 3. A case study to show the importance of Langevin warm-up in the inverse transform that identifies the point set Xb from a mixture representation fb ⋆,σ(ϵ) (i.e., left subfigure). The middle subfigure is the convergence of particles without Langevin warm-up, while the right one is with the warm-up. performance. For example, the last row in the table means not to exclude small groups after clustering converged parti￾cl… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

70 extracted references · 43 canonical work pages

  1. [1]

    Agarwal, R. P. and O'Regan, D. Difference equations in abstract spaces. Journal of the Australian Mathematical Society, 64 0 (2): 0 277--284, 1998

  2. [2]

    The multivariate gaussian probability distribution

    Ahrendt, P. The multivariate gaussian probability distribution. Technical University of Denmark, Tech. Rep, 203, 2005

  3. [3]

    Andrieu, C., De Freitas, N., Doucet, A., and Jordan, M. I. An introduction to mcmc for machine learning. Machine learning, 50: 0 5--43, 2003

  4. [4]

    Single-pass streaming algorithms for correlation clustering

    Behnezhad, S., Charikar, M., Ma, W., and Tan, L.-Y. Single-pass streaming algorithms for correlation clustering. In Proceedings of the 2023 Annual ACM-SIAM Symposium on Discrete Algorithms (SODA), pp.\ 819--849. SIAM, 2023

  5. [5]

    Bilo s , Marin, G. S. Scalable normalizing flows for permutation invariant densities. In International Conference on Machine Learning, pp.\ 957--967. PMLR, 2021

  6. [6]

    Modeling temporal data as continuous functions with stochastic process diffusion

    Bilo s , M., Rasul, K., Schneider, A., Nevmyvaka, Y., and G \"u nnemann, S. Modeling temporal data as continuous functions with stochastic process diffusion. In International Conference on Machine Learning, pp.\ 2452--2470. PMLR, 2023

  7. [7]

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

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

  8. [8]

    Chen, R. T. Q., Amos, B., and Nickel, M. Neural spatio-temporal point processes. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=XQQA6-So14

Show all 70 references
  1. [9]

    and Zhou, M

    Chen, T. and Zhou, M. Learning to jump: Thinning and thickening latent counts for generative modeling. In International Conference on Machine Learning, pp.\ 5367--5382. PMLR, 2023

  2. [10]

    System Data - Citi Bike Trip Histories , 2024

    Citi Bike . System Data - Citi Bike Trip Histories , 2024. URL https://citibikenyc.com/system-data

  3. [11]

    and Kalmykov, Y

    Coffey, W. and Kalmykov, Y. P. The Langevin equation: with applications to stochastic problems in physics, chemistry and electrical engineering, volume 27. World Scientific, 2012

  4. [12]

    and Isham, V

    Cox, D. and Isham, V. Point Processes. Chapman & Hall/CRC Monographs on Statistics & Applied Probability. Taylor & Francis, 1980. ISBN 9780412219108. URL https://books.google.co.uk/books?id=KWF2xY6s3PoC

  5. [13]

    Cox, D. R. Some statistical methods connected with series of events. Journal of the Royal Statistical Society: Series B (Methodological), 17 0 (2): 0 129--157, 1955

  6. [14]

    and Nichol, A

    Dhariwal, P. and Nichol, A. Diffusion models beat gans on image synthesis. Advances in neural information processing systems, 34: 0 8780--8794, 2021

  7. [15]

    M., Kotecha, J

    Djuric, P. M., Kotecha, J. H., Zhang, J., Huang, Y., Ghirmai, T., Bugallo, M. F., and Miguez, J. Particle filtering. IEEE signal processing magazine, 20 0 (5): 0 19--38, 2003

  8. [16]

    Recurrent marked temporal point processes: Embedding event history to vector

    Du, N., Dai, H., Trivedi, R., Upadhyay, U., Gomez-Rodriguez, M., and Song, L. Recurrent marked temporal point processes: Embedding event history to vector. In Proceedings of the 22nd ACM SIGKDD international conference on knowledge discovery and data mining, pp.\ 1555--1564, 2016

  9. [17]

    and Joshi, M

    Friedlander, F. and Joshi, M. Introduction to the Theory of Distributions. Introduction to the Theory of Distributions. Cambridge University Press, 1998. ISBN 9780521649711. URL https://books.google.co.uk/books?id=9O-pXIVUEFMC

  10. [18]

    D., Murphy, K

    Gao, R., Hoogeboom, E., Heek, J., Bortoli, V. D., Murphy, K. P., and Salimans, T. Diffusion meets flow matching: Two sides of the same coin. 2024. URL https://diffusionflow.github.io/

  11. [19]

    Generative adversarial networks

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., and Bengio, Y. Generative adversarial networks. Communications of the ACM, 63 0 (11): 0 139--144, 2020

  12. [20]

    and Graves, A

    Graves, A. and Graves, A. Long short-term memory. Supervised sequence labelling with recurrent neural networks, pp.\ 37--45, 2012

  13. [21]

    M., Rasch, M

    Gretton, A., Borgwardt, K. M., Rasch, M. J., Sch \"o lkopf, B., and Smola, A. A kernel two-sample test. The Journal of Machine Learning Research, 13 0 (1): 0 723--773, 2012

  14. [22]

    Lafma: A latent flow matching model for text-to-audio generation

    Guan, W., Wang, K., Zhou, W., Wang, Y., Deng, F., Wang, H., Li, L., Hong, Q., and Qin, Y. Lafma: A latent flow matching model for text-to-audio generation. In Interspeech 2024, pp.\ 4813--4817, 2024. doi:10.21437/Interspeech.2024-1848

  15. [23]

    Hawkes, A. G. Spectra of some self-exciting and mutually exciting point processes. Biometrika, 58 0 (1): 0 83--90, 1971

  16. [24]

    Denoising diffusion probabilistic models

    Ho, J., Jain, A., and Abbeel, P. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33: 0 6840--6851, 2020

  17. [25]

    Ho, J., Salimans, T., Gritsenko, A., Chan, W., Norouzi, M., and Fleet, D. J. Video diffusion models. Advances in Neural Information Processing Systems, 35: 0 8633--8646, 2022

  18. [26]

    Generating and imputing tabular data via diffusion and flow-based gradient-boosted trees

    Jolicoeur-Martineau, A., Fatras, K., and Kachman, T. Generating and imputing tabular data via diffusion and flow-based gradient-boosted trees. In International Conference on Artificial Intelligence and Statistics, pp.\ 1288--1296. PMLR, 2024

  19. [27]

    Kantorovich, L. V. and Akilov, G. P. Functional analysis. Elsevier, 2014

  20. [28]

    Functional flow matching

    Kerrigan, G., Migliorini, G., and Smyth, P. Functional flow matching. In Dasgupta, S., Mandt, S., and Li, Y. (eds.), Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, volume 238 of Proceedings of Machine Learning Research, pp.\ 3934--3...

  21. [29]

    Neural controlled differential equations for irregular time series

    Kidger, P., Morrill, J., Foster, J., and Lyons, T. Neural controlled differential equations for irregular time series. Advances in Neural Information Processing Systems, 33: 0 6696--6707, 2020

  22. [30]

    Kingma, D. P. and Dhariwal, P. Glow: Generative flow with invertible 1x1 convolutions. Advances in neural information processing systems, 31, 2018

  23. [31]

    Kingman, J. F. C. Poisson processes, volume 3. Clarendon Press, 1992

  24. [32]

    Tabddpm: Modelling tabular data with diffusion models

    Kotelnikov, A., Baranchuk, D., Rubachev, I., and Babenko, A. Tabddpm: Modelling tabular data with diffusion models. In International Conference on Machine Learning, pp.\ 17564--17579. PMLR, 2023

  25. [33]

    Gaussian measures in banach spaces

    Kuo, H.-H. Gaussian measures in banach spaces. Gaussian measures in banach spaces, pp.\ 1--109, 2006

  26. [34]

    A tutorial on energy-based learning

    LeCun, Y., Chopra, S., Hadsell, R., Ranzato, M., Huang, F., et al. A tutorial on energy-based learning. Predicting structured data, 1 0 (0), 2006

  27. [35]

    S., and Hashimoto, T

    Li, X., Thickstun, J., Gulrajani, I., Liang, P. S., and Hashimoto, T. B. Diffusion-lm improves controllable text generation. Advances in Neural Information Processing Systems, 35: 0 4328--4343, 2022

  28. [36]

    B., Azizzadenesheli, K., liu, B., Bhattacharya, K., Stuart, A., and Anandkumar, A

    Li, Z., Kovachki, N. B., Azizzadenesheli, K., liu, B., Bhattacharya, K., Stuart, A., and Anandkumar, A. Fourier neural operator for parametric partial differential equations. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=c8P9NQVtmnO

  29. [37]

    H., Kovachki, N

    Lim, J. H., Kovachki, N. B., Baptista, R., Beckham, C., Azizzadenesheli, K., Kossaifi, J., Voleti, V., Song, J., Kreis, K., Kautz, J., et al. Score-based diffusion models in function space. arXiv preprint arXiv:2302.07400, 2023

  30. [38]

    T., Ben-Hamu, H., Nickel, M., and Le, M

    Lipman, Y., Chen, R. T., Ben-Hamu, H., Nickel, M., and Le, M. Flow matching for generative modeling. arXiv preprint arXiv:2210.02747, 2022

  31. [39]

    u dke, D., Bilo s , M., Shchur, O., Lienen, M., and G \

    L \"u dke, D., Bilo s , M., Shchur, O., Lienen, M., and G \"u nnemann, S. Add and thin: Diffusion for temporal point processes. Advances in Neural Information Processing Systems, 36: 0 56784--56801, 2023

  32. [40]

    u dke, D., Ravent \'o s, E. R., Kollovieh, M., and G \

    L \"u dke, D., Ravent \'o s, E. R., Kollovieh, M., and G \"u nnemann, S. Unlocking point processes through point set diffusion. arXiv preprint arXiv:2410.22493, 2024

  33. [41]

    and Hu, W

    Luo, S. and Hu, W. Diffusion probabilistic models for 3d point cloud generation. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, pp.\ 2837--2845, 2021

  34. [42]

    A conditional point diffusion-refinement paradigm for 3d point cloud completion

    Lyu, Z., Kong, Z., XU, X., Pan, L., and Lin, D. A conditional point diffusion-refinement paradigm for 3d point cloud completion. In International Conference on Learning Representations, 2022. URL https://openreview.net/forum?id=wqD6TfbYkrn

  35. [43]

    Pu-flow: A point cloud upsampling network with normalizing flows

    Mao, A., Du, Z., Hou, J., Duan, Y., Liu, Y.-j., and He, Y. Pu-flow: A point cloud upsampling network with normalizing flows. IEEE Transactions on Visualization and Computer Graphics, 29 0 (12): 0 4964--4977, 2022

  36. [44]

    and Eisner, J

    Mei, H. and Eisner, J. M. The neural hawkes process: A neurally self-modulating multivariate point process. Advances in neural information processing systems, 30, 2017

  37. [45]

    R., and Waagepetersen, R

    M ller, J., Syversveen, A. R., and Waagepetersen, R. P. Log gaussian cox processes. Scandinavian journal of statistics, 25 0 (3): 0 451--482, 1998

  38. [46]

    Gsd: View-guided gaussian splatting diffusion for 3d reconstruction

    Mu, Y., Zuo, X., Guo, C., Wang, Y., Lu, J., Wu, X., Xu, S., Dai, P., Yan, Y., and Cheng, L. Gsd: View-guided gaussian splatting diffusion for 3d reconstruction. In European Conference on Computer Vision, pp.\ 55--72. Springer, 2025

  39. [47]

    Space-time point-process models for earthquake occurrences

    Ogata, Y. Space-time point-process models for earthquake occurrences. Annals of the Institute of Statistical Mathematics, 50: 0 379--402, 1998

  40. [48]

    Deep mixture point processes: Spatio-temporal event prediction with rich contextual information

    Okawa, M., Iwata, T., Kurashima, T., Tanaka, Y., Toda, H., and Ueda, N. Deep mixture point processes: Spatio-temporal event prediction with rich contextual information. In Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, pp.\ 37...

  41. [49]

    Fully neural network based model for general temporal point processes

    Omi, T., Aihara, K., et al. Fully neural network based model for general temporal point processes. Advances in neural information processing systems, 32, 2019

  42. [50]

    On-line new event detection using single pass clustering

    Papka, R., Allan, J., et al. On-line new event detection using single pass clustering. University of Massachusetts, Amherst, 10 0 (290941.290954), 1998

  43. [51]

    Infinite-dimensional diffusion models

    Pidstrigach, J., Marzouk, Y., Reich, S., and Wang, S. Infinite-dimensional diffusion models. arXiv preprint arXiv:2302.10130, 2023

  44. [52]

    On wasserstein two-sample testing and related families of nonparametric tests

    Ramdas, A., Garc \' a Trillos, N., and Cuturi, M. On wasserstein two-sample testing and related families of nonparametric tests. Entropy, 19 0 (2): 0 47, 2017

  45. [53]

    Intensity-free learning of temporal point processes

    Shchur, O., Biloš, M., and Günnemann, S. Intensity-free learning of temporal point processes. In International Conference on Learning Representations, 2020 a . URL https://openreview.net/forum?id=HygOjhEYDH

  46. [54]

    Fast and flexible temporal point processes with triangular maps

    Shchur, O., Gao, N., Bilo s , M., and G \"u nnemann, S. Fast and flexible temporal point processes with triangular maps. Advances in neural information processing systems, 33: 0 73--84, 2020 b

  47. [55]

    Deep unsupervised learning using nonequilibrium thermodynamics

    Sohl-Dickstein, J., Weiss, E., Maheswaranathan, N., and Ganguli, S. Deep unsupervised learning using nonequilibrium thermodynamics. In International conference on machine learning, pp.\ 2256--2265. PMLR, 2015

  48. [56]

    P., Kumar, A., Ermon, S., and Poole, B

    Song, Y., Sohl-Dickstein, J., Kingma, D. P., Kumar, A., Ermon, S., and Poole, B. Score-based generative modeling through stochastic differential equations. In International Conference on Learning Representations, 2021. URL https://openreview.net/forum?id=PxTIG12RRHS

  49. [57]

    Coronavirus (Covid-19) Data in the United States , 2020

    The New York Times . Coronavirus (Covid-19) Data in the United States , 2020. URL https://github.com/nytimes/covid-19-data

  50. [58]

    A note on the evaluation of generative models

    Theis, L., van den Oord, A., and Bethge, M. A note on the evaluation of generative models. In International Conference on Learning Representations, Apr 2016. URL http://arxiv.org/abs/1511.01844

  51. [59]

    Geological Survey

    U.S. Geological Survey . Earthquake Catalogue (accessed August 21, 2020) , 2020. URL https://earthquake.usgs.gov/earthquakes/search/

  52. [60]

    Attention is all you need

    Vaswani, A. Attention is all you need. Advances in Neural Information Processing Systems, 2017

  53. [61]

    Widder, D. V. The heat equation, volume 67. Academic Press, 1976

  54. [62]

    Williams, C. K. and Rasmussen, C. E. Gaussian processes for machine learning, volume 2. MIT press Cambridge, MA, 2006

  55. [63]

    Xie, J., Xu, Y., Zheng, Z., Zhu, S.-C., and Wu, Y. N. Generative pointnet: Deep energy-based learning on unordered point sets for 3d generation, reconstruction and classification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp.\ 14976-...

  56. [64]

    Global convergence of langevin dynamics based algorithms for nonconvex optimization

    Xu, P., Chen, J., Zou, D., and Gu, Q. Global convergence of langevin dynamics based algorithms for nonconvex optimization. Advances in Neural Information Processing Systems, 31, 2018

  57. [65]

    Pointflow: 3d point cloud generation with continuous normalizing flows

    Yang, G., Huang, X., Hao, Z., Liu, M.-Y., Belongie, S., and Hariharan, B. Pointflow: 3d point cloud generation with continuous normalizing flows. In Proceedings of the IEEE/CVF international conference on computer vision, pp.\ 4541--4550, 2019

  58. [66]

    Conditional gan for point cloud generation

    Yang, Z., Chen, Y., Zheng, X., Chang, Y., and Li, X. Conditional gan for point cloud generation. In Proceedings of the Asian Conference on Computer Vision, pp.\ 3189--3205, 2022

  59. [67]

    Spatio-temporal diffusion point processes

    Yuan, Y., Ding, J., Shao, C., Jin, D., and Li, Y. Spatio-temporal diffusion point processes. In Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pp.\ 3173--3184, 2023

  60. [68]

    Self-attentive hawkes process

    Zhang, Q., Lipani, A., Kirnap, O., and Yilmaz, E. Self-attentive hawkes process. In International conference on machine learning, pp.\ 11183--11193. PMLR, 2020

  61. [69]

    Transformer hawkes process

    Zuo, S., Jiang, H., Li, Z., Zhao, T., and Zha, H. Transformer hawkes process. In International conference on machine learning, pp.\ 11692--11702. PMLR, 2020

  62. [70]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 gl...

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.