Pith. sign in

REVIEW 4 major objections 4 minor 30 references

Learning collective multi-cellular dynamics from temporal scRNA-seq via a transformer-enhanced Neural SDE

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Explicitly modeling cell-cell interactions improves reconstruction of held-out gene-expression time points from temporal scRNA-seq data.

desk verdict A genuinely new combination of MV-SDE and transformer attention for temporal scRNA-seq, with strong benchmarks on ZB and MEF; but the claim that the interaction term drives the gains is not isolated by any ablation, and the Panc results undercut the universal statement. read the letter →

arxiv 2505.16492 v1 pith:QBYJRMWS submitted 2025-05-22 q-bio.QM

Qi Jiang , Lei Zhang , Longquan Li , Lin Wan This is my paper · ORCID
classification q-bio.QM
keywords single-cellRNAsequencingcell-cellinteractionsattentionmechanismneuralSDEtransformerMcKean-Vlasovtrajectoryinferencetime-seriesscRNA-seq
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

scIMF is a deep generative model that reconstructs how a population of cells changes over time from snapshot scRNA-seq data, where individual cells are destroyed at each time point. Its central claim is that explicitly modeling how each cell's trajectory is influenced by the rest of the population—rather than treating cells as independent—makes predictions of unseen time points more accurate. The model does this with a neural stochastic differential equation whose drift depends on the population distribution through a transformer self-attention layer, so interaction strengths are learned from gene expression alone. On three published datasets, it reports lower Wasserstein distances than four existing methods on most held-out tasks, with the biggest gains when extrapolating to unobserved late time points.

What carries the argument

The central object is the McKean-Vlasov stochastic differential equation $dx_t = f(x_t, \rho_t)\,dt + \sigma\,dW_t$, where the drift $f(x_t, \rho_t)$ depends on the entire cell-state distribution $\rho_t = \mathrm{Law}(x_t)$. The drift is decomposed into an intrinsic single-cell term $f_{\mathrm{intra}}(x_t)$ and a mean-field interaction term $f_{\mathrm{inter}}(x_t, \rho_t) = \int \phi(x_t, y_t)\,\rho_t(dy_t)$, which is approximated by softmax self-attention over a minibatch of cells. This attention mechanism simultaneously supplies the nonlinear interaction term and yields interpretable, asymmetric cell-to-cell influence scores. The model is trained by optimizing a dynamic optimal transport objective, with a least-action cost regularized by Wasserstein penalties at observed time points, and solved with a NeuralSDE framework.

What would settle it

Run the same held-out tasks with the interaction term ablated, modeling only the cell-intrinsic drift while keeping the architecture and training budget identical; if the full model does not show clearly lower Wasserstein distances than the ablation, the claim that interaction modeling drives the improvement is falsified.

Watch

Extended reading notes

Core claim

The central claim is that adding an explicit, learnable population-dependence term to a single-cell stochastic dynamics model improves the reconstruction of multicellular behavior from temporal scRNA-seq data. scIMF models the cell-state distribution with an interacting mean-field stochastic differential equation, where each cell's drift is the sum of a cell-intrinsic term and an interaction term obtained by integrating pairwise influences over the current population distribution. The interaction term is parameterized by transformer self-attention, so the influence of every other cell on a given cell's motion is learned from data and can be read off as an attention score. On held-out time-point reconstruction tasks across three scRNA-seq datasets, the model reports lower Wasserstein distances than four baselines on most tasks, with the largest gains on extrapolation tasks, and the learned attention matrices are asymmetric, which the authors interpret as non-reciprocal cell-cell communication.

Load-bearing premise

The load-bearing premise is that a cell's fate is influenced by the full population through pairwise, mean-field interactions that the attention mechanism can represent; if real communication is sparse, local, or higher-order, the learned dynamics and the interaction interpretation could be wrong.

Editorial extensions

If this is right

  • Held-out time points, including extrapolated late stages, can be reconstructed from snapshot scRNA-seq data with lower Wasserstein error than current single-cell-level baselines.
  • The estimated attention scores provide time-resolved, directional maps of which cell types influence which, without requiring predefined receptor-ligand pairs.
  • Non-reciprocal interaction patterns can emerge naturally from the asymmetry of learned attention, consistent with out-of-equilibrium multicellular systems.
  • In near-equilibrium in vitro systems, simpler single-cell SDE models may remain sufficient, while interaction-aware modeling matters most for branching in vivo development.

Reading between the lines

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

  • If the attention scores truly reflect biological communication, then comparing them with spatial transcriptomics or lineage-tracing constraints would be a direct validation; the paper does not include such a test.
  • The mean-field integral is approximated from a minibatch of 512 cells, so the learned interaction strengths may depend on sampling; quantifying this sampling variance would clarify how stable the inferred interactions are.
  • The extrapolation gains could come from the interaction term acting as a regularizer on trajectories rather than from capturing true communication; an ablation study would distinguish these explanations.
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 manuscript introduces scIMF, a generative model for temporal scRNA-seq data that combines a McKean-Vlasov stochastic differential equation (MV-SDE) with a transformer-based attention mechanism. The drift is decomposed into a cell-intrinsic term and an interaction term, with the interaction term approximated by softmax attention over minibatches. The model is trained under a dynamic optimal transport objective and evaluated by predicting held-out time points on zebrafish embryogenesis (ZB), mouse embryonic fibroblast reprogramming (MEF), and pancreatic beta-cell differentiation (Panc) data. The authors report improved Wasserstein distances on ZB and MEF held-out tasks relative to PRESCIENT, MIOFlow, PI-SDE, and scNODE, and claim that the attention scores reveal directional, non-reciprocal cell-cell interactions, especially in the ZB data. The paper argues that explicitly modeling cell-cell interactions leads to more accurate reconstructions of multicellular dynamics.

Significance. If the central claim were fully supported, scIMF would be a useful addition to the trajectory-inference toolkit: it targets a genuine limitation of existing methods by incorporating distributional dependence into single-cell dynamics, and it provides a per-cell, time-varying interaction readout. The manuscript also has concrete strengths: the held-out evaluation protocol is a real extrapolation test, the comparisons cover four established baselines, the datasets are public, and the code is released. However, the headline conclusion that explicitly modeling cell-cell interactions is what drives the improved accuracy is not established, because the method differs from every baseline in several coupled design choices and no interaction-free ablation is reported. The Panc results, where PI-SDE outperforms scIMF on every held-out time point, further qualify the universal claim and are addressed only by a post hoc explanation. The biological claims based on attention scores are also not externally validated. These issues are load-bearing because the paper's stated contribution is precisely the interaction mechanism, not merely a new neural-SDE architecture.

major comments (4)
  1. [Methods, Eq. (3)-(4); Results, Tables 2-3; Discussion] The central claim that "explicitly modeling cell-cell interactions leads to more accurate representations of multicellular system dynamics" is not supported by a controlled comparison. scIMF differs from each baseline simultaneously in the transformer attention module, the SDE formulation, and the dynamic optimal transport objective. No ablation is reported in which the interaction term f_inter is disabled, masked, or replaced by a matched-capacity cell-independent function. Without such an ablation, the gains in Tables 2 and 3 could be attributable to the transformer's capacity or to the SDE/OT objective rather than to the mean-field interaction mechanism. I would request an experiment with an scIMF variant in which the attention-based interaction term is removed or replaced (e.g., by a cell-independent MLP of comparable parameter count), evaluated on the same held-out tasks.
  2. [Results, Tables 4-5; Discussion] The universal claim in the Abstract and Introduction is contradicted by the Panc results: PI-SDE achieves the lowest W1 and W2 for all seven held-out time points, while scIMF is second. The paper acknowledges this but explains it post hoc as a quasi-equilibrium regime inferred from the attention matrices themselves. This is circular insofar as the same model outputs are used both to explain the performance gap and to support the biological interpretation. A concrete test would be to quantify the degree of attention asymmetry on Panc versus ZB data and to compare it with an independent measure of equilibrium behavior, or to evaluate whether a symmetric-interaction variant of scIMF matches PI-SDE on Panc. Without such corroboration, the scope of the central claim should be restricted to the ZB and MEF settings.
  3. [Results, Figs. 4-6; Eq. (4)] The biological interpretation of attention scores as cell-cell interaction strengths is not externally validated. The paper uses the model's own attention outputs to claim that interactions are non-reciprocal in ZB data and symmetric in Panc data, but softmax(QK^T/sqrt(d))V is a learned mechanism whose identifiability as a communication strength is not established. I suggest a sanity check: compare attention-based interaction patterns against known receptor-ligand pairs, spatial transcriptomic data, or a perturbation experiment; at minimum, test whether the inferred asymmetry is stable across random seeds and whether a permuted-label or null model produces similar block structure.
  4. [Results, Table 3] The statement that scIMF achieves superior performance in both key evaluation metrics is not uniformly true on MEF data: on the hard task, scIMF's W2 (27.73) is higher than scNODE's (25.98). The text acknowledges this, but the Abstract's phrasing that scIMF "outperforms existing methods in reconstructing gene expression at held-out time points" is too strong. Please qualify the claim to the settings and metrics where the advantage actually holds.
minor comments (4)
  1. [Results, section heading] The text contains typos: "scGenIMF" and "scGeneIMF" should be "scIMF", and "reuslts" should be "results".
  2. [Results, Panc section] The sentence "we conduct held-one-timepoint-out prediction tasksby sequentially holding out" has a missing space in "tasksby"; please correct.
  3. [Methods, Eq. (2)] The notation x_{t0} = rho_{t0} is used, but it would be clearer to write x_{t0} ~ rho_{t0} to indicate that the initial condition is drawn from the distribution, consistent with the later usage in Eq. (9).
  4. [Methods, Optimization] The hyperparameters lambda, sigma, Sinkhorn scaling, and blur are given in the text but are not justified; reporting a brief sensitivity analysis or citing prior work for these choices would improve reproducibility.

Circularity Check

2 steps flagged · score 2.0 of 10

Held-out predictions are genuine extrapolations; minor self-citation in the training objective and a by-construction attention asymmetry do not make the central benchmark circular.

  1. self citation load bearing [STAR Methods, "Solving the interacting diffusion process under the framework of dynamic OT", Eqs. (8)-(9)]
    "This involves ensuring that the modeled diffusion process aligns well with the observed data distribution while adhering to the principle of least action, which governs the natural evolution of biological systems11."

    Reference 11 is the authors' own prior PI-SDE paper (Jiang and Wan, 2024). The scIMF training objective in Eqs. (8)-(9) is essentially the PI-SDE least-action dynamic optimal transport objective, so the central loss function is justified by an appeal to the authors' own prior work rather than by an independent derivation in this paper. This is a modeling choice rather than a fitted prediction, and the held-out benchmark results do not reduce to it; nevertheless, the self-citation is load-bearing for the choice of training objective.

  2. self definitional [Results, "scIMF reveals non-reciprocal cell-cell interactions in ZB dataset"; Methods, Eqs. (4)-(5)]
    "Notably, WQ and WK are trained independently without constraints enforcing their equality. As a result, the estimated attention scores can be asymmetric... Importantly, these attention matrices are non-symmetric, allowing scIMF to capture non-reciprocal relationships between cells. This asymmetry reveals valuable insights into the directionality of cell-cell communication."

    The non-symmetry of the attention matrices is built into the architecture by choosing independent query and key projection matrices with no symmetry constraint. Reporting that the learned attention is non-symmetric and then interpreting this as evidence of biological non-reciprocity is therefore a property inherited from the model definition, not an independent empirical finding. The paper presents the asymmetry as a biological observation, but the design already permits and generically produces asymmetric attention, and no external validation connects attention asymmetry to non-reciprocal cell-cell signaling.

full rationale

The paper's main quantitative claims are genuine held-out extrapolation benchmarks: scIMF is trained on some time points and evaluated on excluded time points via Wasserstein distances, so the reported predictive improvements are not fitted to the evaluation data by construction. The lack of an interaction-free ablation is a substantial validation gap, but it is a correctness concern rather than circularity, because the comparison still tests an independently defined held-out quantity. The two minor circularity issues are (i) the training objective's least-action principle is imported from the authors' own PI-SDE paper, and (ii) the asymmetry of attention scores is presented as a biological discovery when it is a direct consequence of using unconstrained query/key matrices. Neither issue undermines the held-out prediction comparisons, so the overall circularity score is low.

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

The central prediction benchmark rests on standard MV-SDE theory plus the chosen interaction architecture; the biological interpretation adds an unvalidated assumption. The ledger lists the hand-set hyperparameters and modeling choices the quantitative results depend on.

free parameters (5)
  • lambda (Wasserstein penalty weight) = not reported
    In Eq. 9, lambda balances the least-action term and distribution matching; its value is not given or analyzed, and it directly controls the compromise between dynamics and fit.
  • diffusion coefficient sigma = 0.1
    Prescribed constant in STAR Methods; the noise level affects stochastic paths and prediction uncertainty but is not estimated from data.
  • Sinkhorn scaling = 0.7
    Used to compute the Wasserstein loss; chosen by hand with no sensitivity analysis.
  • Sinkhorn blur = 0.1
    Regularization parameter for Sinkhorn; chosen by hand with no sensitivity analysis.
  • PCA dimensionality = 50
    All methods are compared in a 50-dimensional PCA space learned from training data; the choice affects the target space but is applied uniformly.
assumptions (5)
  • standard math The learned drift satisfies conditions for existence and uniqueness of the MV-SDE solution (Eq. 2).
    Relies on Mishura-Veretennikov and Sharrock et al. (refs 27-28); no verification that the trained neural f_theta meets these conditions.
  • ad hoc to paper Cell-cell interactions are additive and pairwise, with f_inter(x_t, rho_t) = integral phi(x_t, y_t) rho_t(dy_t).
    STAR Methods Eq. 3; this is a modeling choice, not derived from biology. The transformer softmax approximates the integral over a minibatch.
  • domain assumption Each observed snapshot is an i.i.d. sample from the population law, and empirical distributions represent rho_t.
    Needed to justify the Wasserstein matching in Eq. 9; ignores batch effects, dropouts, and sampling biases in scRNA-seq.
  • domain assumption The dynamic OT / least-action objective (Eqs. 8-9) yields biologically valid dynamics.
    The objective is inherited from PI-SDE (ref 11, same authors); no independent physical or biological validation is given in this paper.
  • ad hoc to paper Attention scores correspond to real cell-cell communication strengths, and asymmetry implies non-reciprocity.
    Used in the Results to draw biological conclusions from A_ij; no external evidence links attention weights to molecular interactions.
invented entities (1)
  • Asymmetric cell-wise attention scores as cell-cell interaction strengths
    purpose: Quantify directional influence of cell j on cell i and infer cell type communication patterns from attention matrices.
    These scores are internal learned quantities; the paper treats them as biological interactions without validation against receptor-ligand, spatial, or perturbation data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Learning collective multi-cellular dynamics from temporal scRNA-seq via a transformer-enhanced Neural SDE." pith.science (2026). https://pith.science/paper/QBYJRMWS

@misc{pith2026250516492,
  author       = {Pith},
  title        = {Pith review of: Learning collective multi-cellular dynamics from temporal scRNA-seq via a transformer-enhanced Neural SDE},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QBYJRMWS}},
  note         = {Machine review of arXiv:2505.16492}
}
read the original abstract

Time-series single-cell RNA-sequencing (scRNA-seq) datasets offer unprecedented insights into the dynamics and heterogeneity of cellular systems. These systems exhibit multiscale collective behaviors driven by intricate intracellular gene regulatory networks and intercellular interactions of molecules. However, inferring interacting cell population dynamics from time-series scRNA-seq data remains a significant challenge, as cells are isolated and destroyed during sequencing. To address this, we introduce scIMF, a single-cell deep generative Interacting Mean Field model, designed to learn collective multi-cellular dynamics. Our approach leverages a transformer-enhanced stochastic differential equation network to simultaneously capture cell-intrinsic dynamics and intercellular interactions. Through extensive benchmarking on multiple scRNA-seq datasets, scIMF outperforms existing methods in reconstructing gene expression at held-out time points, demonstrating that modeling cell-cell communication enhances the accuracy of multicellular dynamics characterization.Additionally, our model provides biologically interpretable insights into cell-cell interactions during dynamic processes, offering a powerful tool for understanding complex cellular systems.

Figures

Figures reproduced from arXiv: 2505.16492 by the authors.

Figure 1
Figure 1. Overview of scIMF. (A) For time-series data describing the collective cell population dy [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. UMAPs of true and predicted gene expressions for hard held-out task on ZB data (A) [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. scIMF’s performance on ZB data. (A)UMAP of zebrafish embryogenesis cells, colored [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Cell-wise attention scores on ZB data. 8 [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Cell type-wise attention scores on ZB data. The arrow denotes the directional influence [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: The attention scores estimated by scIMF. (A) Cell-wise attention scores on Panc data. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 27 canonical work pages

  1. [1]

    Yuan, G.C., Cai, L., Elowitz, M., Enver, T., Fan, G., Guo, G., Irizarry, R., Kharchenko, P ., Kim, J., Orkin, S. et al. (2017). Challenges and emerging directions in single-cell analysis. Genome biology 18, 1–8

    work page 2017

  2. [2]

    Armingol, E., Officer, A., Harismendy, O., and Lewis, N.E. (2020). Deciphering cell-cell interactions and communication from gene expression. Nature Reviews Genetics 22, 71 – 88

    work page 2020

  3. [3]

    Armingol, E., Baghdassarian, H.M., and Lewis, N.E. (2024). The diversification of methods for studying cell–cell interactions and communication. Nature Reviews Genetics 25, 381 – 400

    work page 2024

  4. [4]

    Kantorovitch, L. (1958). On the translocation of masses. Management science 5, 1–4

    work page 1958

  5. [5]

    Schiebinger, G., Shu, J., Tabaka, M., Cleary, B., Subramanian, V., Solomon, A., Gould, J., Liu, S., Lin, S., Berube, P . et al. (2019). Optimal-transport analysis of single-cell gene expression identifies developmental trajectories in reprogramming. Cell 176, 928–943

    work page 2019

  6. [6]

    Y ang, K.D., and Uhler, C. (2018). Scalable unbalanced optimal transport using generative adversarial networks. arXiv preprint arXiv:1810.11447

    arXiv 2018

  7. [7]

    Benamou, J.D., and Brenier, Y . (2000). A computational fluid mechanics solution to the Monge-Kantorovich mass transfer problem. Numerische Mathematik 84, 375–393

    work page 2000

  8. [8]

    Tong, A., Huang, J., Wolf, G., Van Dijk, D., and Krishnaswamy, S. (2020). Trajectorynet: A dynamic optimal transport network for modeling cellular dynamics. In International Confer- ence on Machine Learning. PMLR pp. 9526–9536

    work page 2020

Show all 30 references
  1. [9]

    Huguet, G., Magruder, D.S., Tong, A., Fasina, O., Kuchroo, M., Wolf, G., and Krishnaswamy, S. (2022). Manifold interpolating optimal-transport flows for trajectory inference. Advances in Neural Information Processing Systems 35, 29705–29718

    2022

  2. [10]

    Y eo, G.H.T., Saksena, S.D., and Gifford, D.K. (2021). Generative modeling of single-cell time series with prescient enables prediction of cell trajectories with interventions. Nature Communications 12, 3222

    2021

  3. [11]

    Jiang, Q., and Wan, L. (2024). A physics-informed neural SDE network for learning cellular dynamics from time-series scRNA-seq data. Bioinformatics 40, ii120–ii127

    2024

  4. [12]

    Zhang, J., Larschan, E., Bigness, J., and Singh, R. (2024). scnode: generative model for temporal single cell transcriptomic data prediction. Bioinformatics 40, ii146–ii154

    2024

  5. [13]

    Jiang, Q., Zhang, S., and Wan, L. (2022). Dynamic inference of cell developmental complex energy landscape from time series single-cell transcriptomic data. PLoS Computational Biology 18, e1009821

    2022

  6. [14]

    Park, C., Mani, S., Beltran-Velez, N., Maurer, K., Huang, T., Li, S., Gohil, S., Livak, K.J., Knowles, D.A., Wu, C.J., and Azizi, E. (2024). A bayesian framework for inferring dynamic intercellular interactions from time-series single-cell data. Genome Research 34, 1384– 1396

    2024

  7. [15]

    Smart, M., and Zilman, A. (2023). Emergent properties of collective gene-expression pat- terns in multicellular systems. Cell Reports Physical Science 4, 101247. 14

    2023

  8. [16]

    Mckean, B.H.P . (1966). A class of markov processes associated with nonlinear parabolic equations. Proceedings of the National Academy of Sciences of the United States of Amer- ica 56, 1907 – 1911

    1966

  9. [17]

    Lacker, D. (2014). A general characterization of the mean field limit for stochastic differential games. Probability Theory and Related Fields 165

    2014

  10. [18]

    Dawson, D.A. (1983). Critical dynamics and fluctuations for a mean-field model of coopera- tive behavior. Journal of Statistical Physics 31, 29–85

    1983

  11. [19]

    Vaswani Ashish, P .N.U.J.J.L.G.A.N.K.A., Shazeer Noam, and Illia, P . (2017). Attention is all you need. In Proceedings of the 31st International Conference on Neural Information Processing Systems 6000–6010

    2017

  12. [20]

    Li, X., Wong, T.K.L., Chen, R.T., and Duvenaud, D. (2020). Scalable gradients for stochastic differential equations. In International Conference on Artificial Intelligence and Statistics. PMLR pp. 3870–3882

    2020

  13. [21]

    Kidger, P ., Foster, J., Li, X., and Lyons, T.J. (2021). Neural sdes as infinite-dimensional gans. In International Conference on Machine Learning. PMLR pp. 5453–5463

    2021

  14. [22]

    Farrell, J.A., Wang, Y ., Riesenfeld, S.J., Shekhar, K., Regev, A., and Schier, A.F . (2018). Single-cell reconstruction of developmental trajectories during zebrafish embryogenesis. Science 360, eaar3131

    2018

  15. [23]

    Schiebinger, G., Shu, J., Tabaka, M., Cleary, B., Subramanian, V., Solomon, A., Gould, J., Liu, S., Lin, S., Berube, P ., Lee, L., Chen, J., Brumbaugh, J., Rigollet, P ., Hochedlinger, K., Jaenisch, R., Regev, A., and Lander, E.S. (2019). Optimal-transport analysis of single-c...

    2019

  16. [24]

    Veres, A., Faust, A.L., Bushnell, H.L., Engquist, E.N., Kenty, J.H.R., Harb, G., Poh, Y .C., Sintov, E., G¨urtler, M., Pagliuca, F .W. et al. (2019). Charting cellular identity during human in vitro β-cell differentiation. Nature 569, 368–373

    2019

  17. [25]

    McInnes, L., Healy, J., and Melville, J. (2018). Umap: Uniform manifold approximation and projection for dimension reduction. arXiv preprint arXiv:1802.03426

    2018 arXiv

  18. [26]

    Fruchart, M., Hanai, R., Littlewood, P .B., and Vitelli, V. (2021). Non-reciprocal phase transi- tions. Nature 592, 363–369

    2021

  19. [27]

    Mishura, Y ., and Veretennikov, A. (2020). Existence and uniqueness theorems for solutions of mckean–vlasov stochastic equations. Theory of Probability and Mathematical Statistics 103, 59–101

    2020

  20. [28]

    Sharrock, L., Kantas, N., Parpas, P ., and Pavliotis, G.A. (2021). Parameter estimation for the McKean-Vlasov stochastic differential equation. arXiv preprint arXiv:2106.13751

    2021 arXiv

  21. [29]

    Y ang, H., Hasan, A., Ng, Y ., and Tarokh, V. (2024). Neural McKean-Vlasov processes: Dis- tributional dependence in diffusion processes. vol. 238 ofProceedings of Machine Learning Research. PMLR pp. 262–270

    2024

  22. [30]

    true McKean-Vlasov case

    Cuturi, M. (2013). Sinkhorn distances: Lightspeed computation of optimal transport. Ad- vances in Neural Information Processing Systems 26. 15 STAR METHODS Data Preprocess We benchmark scIMF against four state-of-the-art methods, PRESCIENT, MIOFlow, PI-SDE, and scNODE, using t...

    2013

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.