Pith. sign in

REVIEW 2 major objections 4 minor 56 references

PyDPF: A Python Package for Differentiable Particle Filtering

T0 review · 2 major / 4 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read PyDPF claims to be the first unified Python package for differentiable particle filters, implementing five gradient-based resampling strategies in one API and validating them by reproducing prior experiments.

desk verdict PyDPF is a real and broadly useful software package, but the printed Sinkhorn/OT algorithm looks inconsistent, so the OT benchmark tables need a code check before they can be trusted. read the letter →

arxiv 2510.25693 v3 pith:HV46LEYQ submitted 2025-10-29 eess.SP cs.LG

John-Joseph Brady , Benjamin Cox , Yunpeng Li , Víctor Elvira This is my paper · ORCID
classification eess.SPcs.LG
keywords differentiableparticlefilteringstate-spacemodelssequentialMonteCarloresamplinggradientestimationparameterlearningsoftwarepackageautomaticdifferentiation
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

Particle filters approximate hidden states in state-space models but do not admit gradients, so parameters cannot be learned by backpropagation. This paper presents PyDPF, a Python package that unifies several recently proposed differentiable particle filters (DPFs)—filters that replace the non-differentiable resampling step with strategies that let gradients flow. The package claims to be the first such implementation with a single API, and validates it on four tasks: a linear-Gaussian model checked against the Kalman filter, stochastic volatility filtering and learning, a visual localisation benchmark, and learning a proposal distribution. If the claim is right, a researcher can load data, define a model, and train or compare DPF variants in a few lines rather than reimplementing each algorithm. The paper's own experiments also indicate which resampling strategies favour low variance versus low bias in practice.

What carries the argument

The load-bearing design is the resampler abstraction: every DPF is a standard particle filter with its resampling step swapped for a differentiable alternative. PyDPF implements five such mechanisms—gradient-truncated resampling, soft resampling, optimal transport resampling, stop-gradient resampling, and kernel mixture resampling—as interchangeable modules with a common interface. Around this, the package supplies cached and constrained parameters for model modules, a dataset/loader convention, an aggregation function to avoid storing all intermediates, and a low-level sequential importance sampling API so users can write custom filters that are not simple resampling swaps. This modularity

What would settle it

Run a built-in DPF and the original published implementation of the same algorithm on identical data, seed, and hyperparameters; if the losses and learned parameters diverge beyond random-seed variation, the package is not faithfully reproducing the algorithm and Section 7's comparisons would not support the paper's conclusions.

Watch

Extended reading notes

Core claim

The central claim is that five differentiable particle filters from the literature can be implemented in one extensible, automatic-differentiation-style package without changing the underlying sequential Monte Carlo structure. PyDPF treats a particle filter as a sequence of importance sampling steps and makes each DPF differ only in its resampler: non-differentiable (gradient-truncated) resampling, soft resampling, optimal transport resampling, stop-gradient resampling, and kernel mixture resampling. The package provides modules for state-space models, datasets, aggregation functions, and a low-level API for custom filters, and it reports experiments reproducing results from previous studies

Load-bearing premise

All benchmark comparisons assume the package's implementations faithfully reproduce the original DPF algorithms, so the reported differences reflect the algorithms themselves rather than quirks or bugs in the code.

Editorial extensions

If this is right

  • Researchers can compare DPF variants on their own state-space models by changing one resampler argument, lowering the overhead of entering this area.
  • The reproduced benchmarks provide a practical map of trade-offs: non-differentiable and soft resampling are fast and low-variance; optimal transport is slow and, in the visual localisation task, hard to stabilise; stop-gradient and marginal variants offer low bias at higher variance.
  • Because the package is extensible, new resampling schemes can be tested against the built-in filters under identical data-loading and training loops, making benchmark results more directly comparable.
  • GPU batching across trajectories is supported, so large-scale parameter learning with DPFs becomes feasible without bespoke parallel code.

Reading between the lines

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

  • A likely knock-on is that DPF research shifts from reimplementing baselines to designing new resamplers, since the package removes the hidden implementation variance that makes literature comparisons unreliable.
  • The empirical dominance of low-variance, biased estimators in the simple parameter-learning example suggests a frontier for future work: estimators that retain the forward pass of a standard filter while reducing the variance of REINFORCE-style gradients, perhaps by better control variates.
  • The package's benchmark harness could be reused to test combinations the paper does not explore, such as learned regularisation strength for optimal transport or learned kernel bandwidth for kernel mixture resampling, which the paper leaves as hyperparameters.
  • The claim of being 'first' is narrower than it sounds: earlier packages cover specific DPF algorithms. The durable contribution is the common API, which makes that claim testable and the field more cumulative.
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

2 major / 4 minor

Summary. The paper presents PyDPF, a Python/PyTorch package that unifies several differentiable particle filters (DPFs) under a common API, specifically implementations of the methods of Jonschkowski et al. (2018), Karkus et al. (2018), Corenflos et al. (2021), Šcibior and Wood (2021), and Younis and Sudderth (2023). The paper describes the package's design, the implemented algorithms, and a set of experiments: linear-Gaussian filtering versus the Kalman filter, stochastic volatility filtering and parameter learning, a DeepMind maze visual-localisation task, and learning proposal parameters. The central claim is that PyDPF is the first unified implementation of such a framework, enabling direct comparison and deployment of multiple DPF algorithms in PyTorch.

Significance. If the implementations are faithful to the original methods, PyDPF is a potentially useful community resource: it provides a single API, documentation, and reproducibility measures (deterministic mode, per-module random generators). The paper is also transparent about known limitations, e.g., it states that no consistency proof exists for kernel-mixture resampling (§5.5). However, the validation is currently weakened by (i) an incorrect printed optimal-transport algorithm that calls into question the fidelity of the OT experiments, and (ii) benchmark tables that mostly omit error bars, making the reported comparisons hard to assess. These issues are load-bearing for the paper's central 'faithful reproduction and benchmarking' claim.

major comments (2)
  1. [§5.3, Algorithms 4–5] The printed optimal-transport resampling algorithm is not a correct Sinkhorn iteration. In Algorithm 4, lines 6–9 loop over n=1,...,d_x, but the potentials f,g are length-K vectors (one entry per particle); in the stochastic volatility experiments d_x=1 and K=100, so 99 of 100 potential entries would never be updated. The updates also swap the marginals: the f-equation uses log(v) and the g-equation uses log(w), whereas the transport plan should couple weight vector w to uniform vector 1/K. Algorithm 5, line 8 sets P_{n,m}=w_n d_x exp(...), with an unexplained factor d_x, and line 11 computes \tilde{X}=d_x P^{(\epsilon)}X; the correct reconstruction implied by Eq. (11) is K (P^{(\epsilon)})^T X. If the code follows the printed algorithm, the OT experiments in Tables 8–12 are not measuring the Corenflos et al. (2021) method; if the code is correct, the text does not describe the implement
  2. [Section 7, Tables 8–12] Several benchmark tables report point estimates without a measure of uncertainty. Table 8 gives a single value per metric; Table 9 gives a gradient s.d. but no variability for α error; Table 10 reports averages over 10 runs without error bars; Table 11 and Table 12 report averages over 5 runs/repeats without standard deviations. Comparative conclusions such as 'soft resampling outperforms stop-gradient' (§7.5) or 'optimal transport is much slower' (§7.2) are not supported without knowing the run-to-run variability. Standard errors or per-seed results should be reported.
minor comments (4)
  1. [§7.2 / Table 8] The reference particle filter is described in the text as having 10,000 particles, while the table caption says 'N=1000 as the reference,' and also lists 'N=100' for the filters. Please reconcile.
  2. [§5.5] Typo 'Summaray' in the summary heading.
  3. [§1] The phrase 'the first implementation of such a framework' is strong given FilterFlow already implements the OT DPF (as acknowledged in §1.1). Recommend qualifying as 'the first unified implementation covering multiple DPF algorithms.'
  4. [§5.3] Please specify the Sinkhorn stopping criterion exactly (norm of potential updates? all batches?) and the decay schedule for \epsilon.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: PyDPF is an implementation and benchmarking package whose claims do not reduce to its own inputs.

full rationale

The paper makes no first-principles theoretical prediction whose output is defined by its inputs. Its central claim is that PyDPF is a unified implementation of already-published differentiable particle filters (Jonschkowski et al. 2018; Karkus et al. 2018; Corenflos et al. 2021; Šcibior and Wood 2021; Younis and Sudderth 2023). Each implementation is anchored to an external algorithm: e.g., Section 5.3 states, 'Our implementation of the Sinkhorn loop in Alg. 4 is a PyTorch reimplementation of FilterFlow’s (Corenflos et al. 2021)', and Section 7 validates against external references such as a Kalman filter, a high-particle-count SIR filter, and published experimental setups. These are external benchmarks, not fitted parameters renamed as predictions. The paper does contain several self-citations (Brady et al. 2025; Chen and Li 2024; Cox et al. 2024; Li et al. 2024), but they are used as prior work or to justify experimental choices, not as load-bearing evidence that PyDPF is correct or as a uniqueness argument forcing the paper's conclusions; the central package functionality is independently implementable from the cited external literature. One potentially serious issue surfaced in review is not circularity: the printed optimal-transport algorithms (Algs. 4-5) appear dimensionally inconsistent with FilterFlow (Sinkhorn updates loop over d_x rather than the particle count K, and the reconstruction is written as tilde-X = d_x P^(epsilon) X rather than the transport-map form K (P^(epsilon))^T X). If the shipped code follows the printed text, this would be an implementation-fidelity or correctness problem for the OT benchmarks, not a reduction of the paper's claims to its own inputs. No circular step is therefore established.

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

The central claim is a software artifact, so there are no scientific free parameters or invented entities. The hyperparameters listed are user-chosen settings in the experiments, not fitted to make the package work. The axioms are standard background assumptions about SMC theory and PyTorch's gradient engine.

free parameters (4)
  • soft resampling ξ = 0.7 in experiments
    Chosen by hand; controls bias-variance trade-off; different values likely change comparative results.
  • OT regularisation ε = 0.5 in experiments
    Chosen by hand; ideally scales with 1/log N; affects stability and Monte Carlo error.
  • kernel bandwidth (kernel-mixture) = 0.1 in Table 8, 0.3 in Tables 9-10
    Chosen by hand; no tuning procedure described; affects gradient variance and accuracy.
  • OT hyperparameters (decay_rate, min_update_size, max_iterations, transport_gradient_clip) = 0.9, 0.01, 100, 1.0
    Followed FilterFlow defaults; treated as fixed, not learned; affect convergence and stability.
assumptions (3)
  • domain assumption Particle filters/SMC provide consistent estimates of filtering distributions and likelihoods.
    Standard SMC theory invoked in Section 3.2/3.3 (Doucet et al. 2009; Andrieu et al. 2010).
  • domain assumption PyTorch autodiff correctly computes gradients through the implemented differentiable resampling operations.
    The package's functionality relies on PyTorch's automatic differentiation; the paper notes determinism caveats in Section 2.4.
  • domain assumption The referenced DPF algorithms are as described in their original papers and their theoretical properties (e.g., OT consistency in Corenflos et al. 2021) hold.
    PyDPF implements these methods; if the original descriptions are wrong, PyDPF inherits the error.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PyDPF: A Python Package for Differentiable Particle Filtering." pith.science (2026). https://pith.science/paper/HV46LEYQ

@misc{pith2026251025693,
  author       = {Pith},
  title        = {Pith review of: PyDPF: A Python Package for Differentiable Particle Filtering},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HV46LEYQ}},
  note         = {Machine review of arXiv:2510.25693}
}
read the original abstract

State-space models (SSMs) are a widely used tool in time series analysis. In the complex systems that arise from real-world data, it is common to employ particle filtering (PF), an efficient Monte Carlo method for estimating the hidden state corresponding to a sequence of observations. Applying particle filtering requires specifying both the parametric form and the parameters of the system, which are often unknown and must be estimated. Gradient-based optimisation techniques cannot be applied directly to standard particle filters, as the filters themselves are not differentiable. However, several recently proposed methods modify the resampling step to make particle filtering differentiable. In this paper, we present an implementation of several such differentiable particle filters (DPFs) with a unified API built on the popular PyTorch framework. Our implementation makes these algorithms easily accessible to a broader research community and facilitates straightforward comparison between them. We validate our framework by reproducing experiments from several existing studies and demonstrate how DPFs can be applied to address several common challenges with state space modelling.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

56 extracted references · 2 canonical work pages

  1. [1]

    Particle M arkov Chain M onte C arlo Methods

    Andrieu C, Doucet A, Holenstein R (2010). Particle M arkov Chain M onte C arlo Methods. Journal of the Royal Statistical Society Series B: Statistical Methodology, 72(3), 269--342

    2010

  2. [2]

    Beattie C, Leibo JZ, Teplyashin D, Ward T, Wainwright M, K \"u ttler H, Lefrancq A, Green S, Vald \'e s V, Sadik A, et al. (2016). Deep M ind Lab. arXiv preprint arXiv:1612.03801

    arXiv 2016

  3. [3]

    Interacting Multiple Model Particle Filter

    Boers Y, Driessen J (2003). Interacting Multiple Model Particle Filter. IEE Proc. Radar, Sonar Nav., 150, 344--349. ISSN 1350-2395

    2003

  4. [4]

    JAX : Composable Transformations of Python + NumPy Programs

    Bradbury J, Frostig R, Hawkins P, Johnson MJ, Leary C, Maclaurin D, Necula G, Paszke A, Vander P las J, Wanderman- M ilne S, Zhang Q (2018). JAX : Composable Transformations of Python + NumPy Programs. ://github.com/jax-ml/jax

    2018

  5. [5]

    Differentiable Interacting Multiple Model Particle Filtering

    Brady JJ, Luo Y, Wang W, Elvira V, Li Y (2025). Differentiable Interacting Multiple Model Particle Filtering. Signal Processing, 238, 110166

    2025

  6. [6]

    LowLevelParticleFilters.jl

    Carlson FB (2025). LowLevelParticleFilters.jl . ://github.com/baggepinnen/LowLevelParticleFilters.jl

    2025

  7. [7]

    Improved Particle Filter for Nonlinear Problems

    Carpenter J, Clifford P, Fearnhead P (1999). Improved Particle Filter for Nonlinear Problems. In IEE Proc. Radar, Sonar and Navi., volume 146

    1999

  8. [8]

    Tracking Measles Infection through Non-Linear State Space Models

    Chen S, Fricks J, Ferrari MJ (2011). Tracking Measles Infection through Non-Linear State Space Models. Journal of the Royal Statistical Society Series C: Applied Statistics, 61(1), 117--134. ISSN 0035-9254. doi:10.1111/j.1467-9876.2011.01001.x. https://academic.oup.com/jrsssc/article-pdf/61/1/117/49548553/jrsssc\_61\_1\_117.pdf , ://doi.org/10.1111/j.1467...

    arXiv 2011

Show all 56 references
  1. [9]

    Normalizing Flow-Based Differentiable Particle Filters

    Chen X, Li Y (2024). Normalizing Flow-Based Differentiable Particle Filters. IEEE Transactions on Signal Processing

    2024

  2. [10]

    An Introduction to Sequential M onte C arlo , chapter Particle Filtering, pp

    Chopin N, Papaspiliopoulos O (2020). An Introduction to Sequential M onte C arlo , chapter Particle Filtering, pp. 129--165. Springer

    2020

  3. [11]

    Operational Implementation of a Hybrid Ensemble/4D- V ar Global Data Assimilation System at the M et O ffice

    Clayton AM, Lorenc AC, Barker DM (2013). Operational Implementation of a Hybrid Ensemble/4D- V ar Global Data Assimilation System at the M et O ffice. Quarterly Journal of the Royal Meteorological Society, 139(675), 1445--1461

    2013

  4. [12]

    Differentiable Particle Filtering via Entropy-Regularized Optimal Transport

    Corenflos A, Thornton J, Deligiannidis G, Doucet A (2021). Differentiable Particle Filtering via Entropy-Regularized Optimal Transport. In Proc. Int. Conf. on Machine Learn. (ICML), pp. 2100--2111. Online

    2021

  5. [13]

    End-to-end Learning of G aussian Mixture Proposals using Differentiable Particle Filters and Neural Networks

    Cox B, Pérez-Vieites S, Zilberstein N, Sevilla M, Segarra S, Elvira V (2024). End-to-end Learning of G aussian Mixture Proposals using Differentiable Particle Filters and Neural Networks. In Int. Conf. Acoustics, Speech and Sig. Proc. (ICASSP), pp. 9701--9705. doi:10.1109/ICAS...

    2024

  6. [14]

    Sinkhorn Distances: Lightspeed Computation of Optimal Transport

    Cuturi M (2013). Sinkhorn Distances: Lightspeed Computation of Optimal Transport. Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 26

    2013

  7. [15]

    Doucet A, Johansen AM, et al. (2009). A Tutorial on Particle Filtering and Smoothing: Fifteen Years Later. Handbook of nonlinear filtering, 12(656-704), 3

    2009

  8. [16]

    Elucidating the Auxiliary Particle Filter via Multiple Importance Sampling

    Elvira V, Martino L, Bugallo MF, Djuri\' c PM (2019). Elucidating the Auxiliary Particle Filter via Multiple Importance Sampling. IEEE Sig. Process. Magazine, 36(6), 145--152

    2019

  9. [17]

    Foerster J, et al. (2018). D i CE : The Infinitely Differentiable M onte C arlo Estimator. In Proc. Int. Conf. on Machine Learning (ICML), volume 80

    2018

  10. [18]

    Turing : a Language for Flexible Probabilistic Inference

    Ge H, Xu K, Ghahramani Z (2018). Turing : a Language for Flexible Probabilistic Inference. In International Conference on Artificial Intelligence and Statistics, AISTATS 2018, 9-11 April 2018, Playa Blanca, Lanzarote, Canary Islands, Spain , pp. 1682--1690. ://proceedings.mlr....

    2018

  11. [19]

    Novel Approach to Nonlinear and Non- G aussian B ayesian State Estimation

    Gordon N, Salmond D, Smith AFM (1993). Novel Approach to Nonlinear and Non- G aussian B ayesian State Estimation. IEE Proceedings-F Radar and Signal Processing, 140, 107--113

    1993

  12. [20]

    Hoffman MD, Gelman A, et al. (2014). The No-U-Turn Sampler: Adaptively Setting Path Lengths in H amiltonian M onte C arlo. J. Mach. Learn. Res., 15(1), 1593--1623

    2014

  13. [21]

    Differentiable Particle Filters: End-to-End Learning with Algorithmic Priors

    Jonschkowski R, Rastogi D, Brock O (2018). Differentiable Particle Filters: End-to-End Learning with Algorithmic Priors. In Proc. Robot.: Sci. Syst. Pittsburgh, PA, USA

    2018

  14. [22]

    A New Approach to Linear Filtering and Prediction Problems

    K alman RE (1960). A New Approach to Linear Filtering and Prediction Problems. Transactions of the ASME--Journal of Basic Engineering, 82(Series D), 35--45

    1960

  15. [23]

    On Particle Methods for Parameter Estimation in State-Space Models

    Kantas N, Doucet A, Singh SS, Maciejowski J, Chopin N (2015). On Particle Methods for Parameter Estimation in State-Space Models. Stat. Sci., pp. 328--351

    2015

  16. [24]

    Particle Filter Networks with Application to Visual Localization

    Karkus P, Hsu D, Lee WS (2018). Particle Filter Networks with Application to Visual Localization. In Proc. Conf. Robot Learn., pp. 169--178. PMLR, Zurich, CH

    2018

  17. [25]

    pomp : Statistical Inference for Partially Observed M arkov Processes

    King AA, Ionides EL, Bret\'o CM, Ellner SP, Ferrari MJ, Funk S, Johnson SG, Kendall BE, Lavine M, Nguyen D, O'Dea EB, Reuman DC, Wearing H, Wood SN (2025). pomp : Statistical Inference for Partially Observed M arkov Processes . doi:10.5281/zenodo.15364462. R package, version 6...

    2025 doi

  18. [26]

    Statistical Inference for Partially Observed M arkov Processes via the R Package pomp

    King AA, Nguyen D, Ionides EL (2016). Statistical Inference for Partially Observed M arkov Processes via the R Package pomp . Journal of Statistical Software, 69(12), 1–43. doi:10.18637/jss.v069.i12. ://www.jstatsoft.org/index.php/jss/article/view/v069i12

    2016 doi

  19. [27]

    Adam: A Method for Stochastic Optimization

    Kingma DP, Ba J (2014). Adam: A Method for Stochastic Optimization. arXiv preprint arXiv:1412.6980

    2014 arXiv

  20. [28]

    Auto-Encoding Variational Bayes

    Kingma DP, Welling M (2013). Auto-Encoding Variational Bayes. arXiv preprint arXiv:1312.6114

    2013 arXiv

  21. [29]

    Toward Practical N^2 M onte C arlo: the Marginal Particle Filter

    Klaas M, de Freitas N, Doucet A (2005). Toward Practical N^2 M onte C arlo: the Marginal Particle Filter. In Proc. Conf. Uncert. Art. Intell. (UAI), pp. 308--315. Arlington, Virginia

    2005

  22. [30]

    Auto-Encoding Sequential M onte C arlo

    Le T, Igl M, Rainforth T, Jin T, Wood F (2018). Auto-Encoding Sequential M onte C arlo. In Proc. Int. Conf. Learn Represent. (ICLR). Vancouver, Canada

    2018

  23. [31]

    An Analysis of Regularized Interacting Particle Methods for Nonlinear Filtering

    LeGland F, Musso C, Oudjane N (1998). An Analysis of Regularized Interacting Particle Methods for Nonlinear Filtering. In IEEE Euro. Works. Computer Intensive Methods in Control and Data Process., pp. 167--174. Prague, Czech Republic

    1998

  24. [32]

    Revisiting Semi-Supervised Training Objectives for Differentiable Particle Filters

    Li J, Brady JJ, Chen X, Li Y (2024). Revisiting Semi-Supervised Training Objectives for Differentiable Particle Filters. In 2024 IEEE 13rd Sensor Array and Multichannel Signal Processing Workshop (SAM), pp. 1--5. IEEE

    2024

  25. [33]

    Particle Gibbs with Ancestor Sampling

    Lindsten F, Jordan MI, Sch \"o n TB (2014). Particle Gibbs with Ancestor Sampling. The Journal of Machine Learning Research, 15(1), 2145--2184

    2014

  26. [34]

    Maddison CJ, et al. (2017). Filtering Variational Objectives. In Proc. Adv. in Neural Info. Process. Syst. (NeurIPS). Long Beach, CA, USA

    2017

  27. [35]

    MATLAB Control System Toolbox

    MathWorks Inc (2025). MATLAB Control System Toolbox . ://uk.mathworks.com/help/control/index.html

    2025

  28. [36]

    M onte C arlo Gradient Estimation in Machine Learning

    Mohamed S, Rosca M, Figurnov M, Mnih A (2020). M onte C arlo Gradient Estimation in Machine Learning. J. Mach. Learn. Research, 21(132)

    2020

  29. [37]

    Feynman- K ac Formulae: Genealogical and Interacting Particle Systems with Applications

    Moral P (2004). Feynman- K ac Formulae: Genealogical and Interacting Particle Systems with Applications . Springer

    2004

  30. [38]

    pypfilt : a Particle Filter for Python

    Moss R (2024). pypfilt : a Particle Filter for Python. Journal of Open Source Software, 9(96), 6276. doi:10.21105/joss.06276. ://doi.org/10.21105/joss.06276

    2024 doi

  31. [39]

    Improving Regularized Particle Filters

    Musso C, Oudjane N, Le Grand F (2001). Improving Regularized Particle Filters. In A Doucet, N Freitas, N Gordon (eds.), Sequential M onte C arlo Methods in Practice , pp. 247--271. Springer-Verlang

    2001

  32. [40]

    Variational Sequential M onte C arlo

    Naesseth C, Linderman S, Ranganath R, Blei D (2018). Variational Sequential M onte C arlo. In Proc. Int. Conf. Art. Int. and Stat. (AISTATS), pp. 968--977. PMLR, Lanzarote, Canary Islands

    2018

  33. [41]

    Neal RM, et al. (2011). MCMC using Hamiltonian Dynamics. Handbook of M arkov chain M onte C arlo , 2(11), 2

    2011

  34. [42]

    A Simplex Method for Function Minimization

    Nelder JA, Mead R (1965). A Simplex Method for Function Minimization. The computer journal, 7(4), 308--313

    1965

  35. [43]

    State-Space Models for Ecological Time-Series Data: Practical Model-Fitting

    Newman K, King R, Elvira V, de Valpine P, McCrea RS, Morgan BJ (2023). State-Space Models for Ecological Time-Series Data: Practical Model-Fitting. Methods in Ecology and Evolution, 14(1), 26--42

    2023

  36. [44]

    Variational Bayesian inference with stochastic search

    Paisley J, Blei DM, Jordan MI (2012). Variational Bayesian inference with stochastic search. In Proc. Int. Conf. Machine Learning (ICML), pp. 1363--1370

    2012

  37. [45]

    Normalizing Flows for Probabilistic Modeling and Inference

    Papamakarios G, Nalisnick E, Rezende DJ, Mohamed S, Lakshminarayanan B (2021). Normalizing Flows for Probabilistic Modeling and Inference. J. of Mach. Learn. Research, 22(57), 1--64

    2021

  38. [46]

    PyTorch : An Imperative Style, High-Performance Deep Learning Library

    Paszke A (2019). PyTorch : An Imperative Style, High-Performance Deep Learning Library. arXiv preprint arXiv:1912.01703

    2019 arXiv

  39. [47]

    Filtering via Simulation: Auxiliary Particle Filters

    Pitt MK, Shephard N (1999). Filtering via Simulation: Auxiliary Particle Filters. Journal of the American statistical association, 94(446), 590--599

    1999

  40. [48]

    Bayesian Filtering and Smoothing, volume 17

    S \"a rkk \"a S, Svensson L (2023). Bayesian Filtering and Smoothing, volume 17. Cambridge university press

    2023

  41. [49]

    Differentiable Particle Filtering Without Modifying the Forward Pass

    \'S cibior A, Wood F (2021). Differentiable Particle Filtering Without Modifying the Forward Pass. arXiv:2106.10314

    2021 arXiv

  42. [50]

    Particle Learning for B ayesian Semi-Parametric Stochastic Volatility Model

    Virbickaite A, Lopes HF, Ausin MC, Galeano P (2019). Particle Learning for B ayesian Semi-Parametric Stochastic Volatility Model. Econometric Reviews

    2019

  43. [51]

    A Survey of Recent Advances in Particle Filters and Remaining Challenges for Multitarget Tracking

    Wang X, Li T, Sun S, Corchado JM (2017). A Survey of Recent Advances in Particle Filters and Remaining Challenges for Multitarget Tracking. Sensors, 17(12), 2707

    2017

  44. [52]

    Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning

    Williams RJ (1992). Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning. Machine Learning, 8, 229--256

    1992

  45. [53]

    Differentiable and Stable Long-Range Tracking of Multiple Posterior Modes

    Younis A, Sudderth E (2023). Differentiable and Stable Long-Range Tracking of Multiple Posterior Modes. In Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), volume 36. New Orleans, LA, USA

    2023

  46. [54]

    Learning to be Smooth: An End-to-End Differentiable Particle Smoother

    Younis A, Sudderth E (2024). Learning to be Smooth: An End-to-End Differentiable Particle Smoother. Adv. Neural Inf. Process. Syst., 37, 7125--7155

    2024

  47. [55]

    , " * write output.state after.block = add.period write newline

    ENTRY address archive author booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key month note number numpages organization pages publisher school series title type url volume year label extra.label sort.label short.list IN...

  48. [56]

    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 4, 2026 · model on record in the stance chip above.