Pith. sign in

REVIEW 4 major objections 4 minor 21 references

Ensemble Kalman filter for uncertainty in human language comprehension

T0 review · 4 major / 4 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read A Bayesian last layer makes a sentence model hedge when syntax and meaning conflict.

desk verdict A new application of an EnKF sampler to a small sentence-processing model yields a real but prior-sensitive change in predictions; the Bayesian and human-uncertainty claims need more support. read the letter →

arxiv 2505.02590 v1 pith:SS5UVQHD submitted 2025-05-05 cs.CL stat.APstat.ML

Diksha Bhandari , Alessandro Lopopolo , Milena Rabovsky , Sebastian Reich This is my paper · ORCID
classification cs.CLstat.APstat.ML MSC 62F1565C3568T07
keywords BayesianinferenceensembleKalmanfilteruncertaintyquantificationsentencecomprehensionGestaltmodelreversalanomalylast-layerinteractingparticlesystems
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 asks whether a neural network model of sentence comprehension can represent uncertainty instead of committing to a single reading. The authors take the Sentence Gestalt model, train it with maximum likelihood, and then apply an ensemble-Kalman-filter-based Bayesian sampler to its output layer, comparing the two versions on reversal anomalies—sentences in which syntactic role assignment and semantic plausibility point in opposite directions. They find that the standard model confidently keeps the semantically plausible but syntactically wrong reading, while the Bayesian model lowers activation for that reading and raises activation for the syntactically indicated roles. They argue that this graded, cue-sensitive response is closer to the uncertainty humans show on such sentences, which behavioral studies link to errors and reanalysis. If the finding holds, last-layer Bayesian inference offers a practical route to uncertainty estimates in existing comprehension models without retraining their core.

What carries the argument

The central mechanism is the dropout deterministic sampler, an interacting-particle-system ordinary differential equation that evolves an ensemble of output-layer weight vectors from a Gaussian prior toward the posterior in the limit of large simulation time. It is an extension of the ensemble Kalman filter adapted to logistic-regression-style losses, and it is applied only to the last layer of the Sentence Gestalt model after the earlier layers have been fixed by maximum-likelihood training. Dropout randomly zeroes entries of the ensemble to prevent the particles from collapsing into the low-dimensional subspace spanned by their initial values, and a Trotter-splitting time-stepping scheme keeps the numerical cost low. The final ensemble's averaged sigmoid predictions replace the single point estimate in the predictive distribution.

What would settle it

Run the same eight reversal-anomaly sentences with human participants and measure a behavioral or neural index of uncertainty, such as error rate, reaction time, or the late positive brain response associated with reanalysis. If the Bayesian model's activation shift appears on items where humans show no uncertainty, or fails to appear where humans do show uncertainty, the claimed correspondence between model activations and human processing would be refuted. A computational check would be to vary the prior covariance over a wide range; the paper's own appendix shows a very small prior covariance removes the effect.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that replacing the point estimate of the Sentence Gestalt model's output layer with a posterior over weights changes how the model handles reversal anomalies. For semantically congruent sentences, the Bayesian model's output activations are comparable to those of the maximum-likelihood model. For reversal anomalies, the MLE model continues to activate the semantically plausible but syntactically wrong role assignments with high confidence, whereas the Bayesian model's mean activations for those assignments drop and its activations for the syntactically indicated roles rise. Statistical tests in the paper show that the Bayesian activations for reversed words differ significantly from zero and from the MLE model's activations, while the MLE model's do not; the authors interpret this as the model representing graded uncertainty when semantic and syntactic cues conflict, and take it as closer to human processing.

Load-bearing premise

The load-bearing premise is that changes in the Bayesian model's output activations correspond to the uncertainty humans experience on reversal anomalies; that mapping is borrowed from earlier behavioral and ERP studies and is not directly tested with human data in this paper, and the posterior itself is defined relative to a Gaussian prior centered at the maximum-likelihood solution.

Editorial extensions

If this is right

  • A last-layer Bayesian treatment can be added to an already maximum-likelihood-trained sentence model, so uncertainty estimates do not require retraining the feature layers.
  • On reversal anomalies, the model no longer commits fully to the semantically plausible reading; both the syntactic and the semantic candidate roles remain represented with graded activation.
  • The same activation shift does not appear for irrelevant role assignments, indicating that the increased uncertainty tracks the actual semantic-syntactic conflict rather than general noise.
  • Predictive probabilities, not just point estimates, become available for any probe question, offering a route to calibrated confidence in sentence-processing models.

Reading between the lines

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

  • A direct item-level experiment, not run in the paper, would compare the Bayesian model's activation differences with human reaction times, error rates, or ERP amplitudes on the same sentences; that would test whether the borrowed human-uncertainty mapping actually holds.
  • Because the prior is centered at the maximum-likelihood solution and the paper's Appendix D shows the effect vanishes for a very small prior covariance, some of the reported uncertainty may be a prior-width artifact rather than purely data-driven posterior behavior.
  • Only the output layer is treated Bayesian here; uncertainty in the learned feature representation itself is not quantified, so genuinely ambiguous semantic representations might require a deeper Bayesian treatment.
  • The same EnKF-based sampler could in principle be applied to output layers of other layered cognitive or language models as a cheap uncertainty signal, though the frozen-feature assumption would need to be checked in each setting.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper proposes a Bayesian extension of the Sentence Gestalt (SG) model of sentence comprehension, using an ensemble Kalman filter (EnKF) based dropout deterministic sampler to perform last-layer Bayesian inference. The SG model is first trained by maximum likelihood estimation (MLE), the feature map is frozen, and a Gaussian prior is placed on the output-layer parameters; predictions are obtained by averaging sigmoid outputs over the final particle ensemble (Eq. 5). The authors compare this Bayesian SG model with the MLE-trained model on semantically congruent and reversal-anomaly (RA) sentences, reporting that the Bayesian model lowers activations for semantically plausible but syntactically incorrect role assignments and raises activations for syntactically indicated roles. This behavior is interpreted as greater uncertainty under conflicting cues and as being closer to human comprehension. The evidence consists of model-level and item-level t-tests on simulated test sentences; no human behavioral or neurophysiological data are collected.

Significance. If the central claim were established, the paper would make a useful methodological contribution by showing that a relatively cheap last-layer Bayesian correction can change the qualitative behavior of a cognitive model under cue conflict, and the deterministic EnKF sampler is of independent interest. The paper is transparent about its hyperparameters, provides pseudocode, and includes a prior-sensitivity appendix, which are strengths. However, the central empirical claim is not yet supported: the prior is centered at the MLE estimated on the same training data, the reversal-anomaly effect largely disappears under a smaller prior covariance (Appendix D), and no human data directly test the claim that the Bayesian model is closer to human processing. These issues are load-bearing for the paper's stated conclusions rather than cosmetic.

major comments (4)
  1. [Section 3, Eq. (1)] The prior is data-dependent in a way that undermines the Bayesian interpretation. The manuscript states that a Gaussian prior with mean mprior = θMLE and covariance Pprior = I is used, where θMLE is obtained by training the full SG model on the same dataset D (Section 2.4). Because the prior mean is a function of the same labels that appear in the likelihood, Eq. (1) does not define a posterior under any pre-data belief. Consequently the ensemble spread and the predictive distribution in Eq. (5) do not measure epistemic uncertainty in the standard Bayesian sense; they reflect a mixture of the likelihood and an ad hoc, data-dependent regularization. This is load-bearing because the central claim of the paper is about uncertainty representation. The authors should either use a prior that is genuinely independent of D (e.g., a zero-mean prior with a justified covariance) or explicitly reframe the method as a regularized ensemble procedure and avoid the language of Bayesian posterior uncertainty.
  2. [Appendix D, Table D.8] The main qualitative result is controlled by an arbitrary hyperparameter. The paired t-tests in Table D.8 show that with Pprior = 0.01I, activations from the dropout deterministic sampler are not significantly different from the ADAM-trained model (p > 0.05), whereas with Pprior = I and Pprior = 5I they are highly significant. This means the reported 'increased uncertainty' on reversal anomaly sentences is not a robust consequence of the data or of Bayesian inference; it is a consequence of choosing a sufficiently large prior covariance. Because this effect is exactly what the paper uses to support its main conclusion, the authors need to provide a principled justification for the prior scale or demonstrate that the qualitative pattern holds across a range of defensible priors. As it stands, Appendix D documents that the central empirical claim is prior-driven rather than data-driven.
  3. [Section 4, Discussion] The claim that the Bayesian SG model is 'closer to human cognitive processing' is not tested against any human data. The paper reports only model-model differences on synthetic test sentences; the mapping from output activations to human uncertainty is qualitative and is borrowed from cited behavioral and ERP studies (Ferreira 2003; Kuperberg et al. 2003). No behavioral accuracy, reaction time, or ERP benchmark is used to validate the proposed model. Even if the Bayesian model shows reduced confidence on RA sentences, this by itself does not establish that it approximates human comprehension. The authors should either add a direct comparison with human behavioral or ERP data, or explicitly limit the conclusion to 'increased activation uncertainty' in the model and remove or substantially weaken the human-approximation claim.
  4. [Eq. (5), Section 3.1] Part of the reported 'greater uncertainty' is built in by the ensemble averaging formula. Equation (5) averages sigmoid outputs over the particle ensemble, so any ensemble with positive covariance will smooth extreme activations relative to a single point estimate. Since the ensemble spread is governed by the prior covariance, the difference between the Bayesian and ADAM activations in Figures 4 and 5 is at least partly inherited from the prior rather than learned from the data. This is related to the prior-sensitivity problem above, but it deserves its own statement because it means the qualitative pattern is not evidence of a specific cognitive mechanism unless the prior is justified independently.
minor comments (4)
  1. [Table 3 and Figure 5] There appears to be an inconsistency between the t-test results and the figure: Table 3 reports that ADAM activations for reversed words do not differ significantly from zero, yet Figure 5 shows high mean activations for the semantically plausible roles. Please clarify what quantity is being tested, correct any mismatch between the table labels and the figure legends, and report the mean and standard deviation for the tested units.
  2. [Abstract and Section 2.1.1] There are several typos: 'extention' in the abstract, 'determinsitic' in Section 2.1.1, and 'the the' in Section 3.1. These should be corrected before publication.
  3. [Figures 4 and 5] The figure captions refer to standard deviations, but the text reports t-tests over 10 runs and item-level analyses over 8 items. Please specify whether the error bars represent variability across models, across items, or across both, and state which analysis is displayed.
  4. [Appendix D] The significance threshold is used inconsistently: the main text uses p < 0.05, while Table D.8 uses p < 0.001. Please use a consistent alpha level and state it explicitly.

Circularity Check

2 steps flagged · score 6.0 of 10

The reported 'greater uncertainty' on reversal anomalies is largely built into the last-layer Bayesian construction: a posterior predictive that averages sigmoids with a data-dependent MLE-centered prior will soften extreme activations by construction, and Appendix D shows the effect vanishes when the prior covariance is small.

  1. fitted input called prediction [Section 3, paragraph 'Training using the MLE approach...' (also Sections 2.4 and 2.2)]
    "We first train the complete SG model using a regularized MLE technique as described in Section 2.4. ... This feature map ψMLE(x) is used for Bayesian logistic regression over multiple output features. We implement the proposed dropout deterministic sampler for Bayesian inference using a Gaussian prior with mean mprior = θMLE and covariance matrix Pprior = I."

    The prior in Eq. (1) is introduced as a belief held before observing training data ('we put a prior distribution π(θ) over the parameter space, which represents our belief as to which parameters are likely to have generated our output before our model observes any training data'). Here, however, the prior mean is the very MLE obtained by fitting the same dataset D that enters the likelihood l(θ) in Eq. (1). The resulting 'posterior' is therefore not a posterior under a pre-data belief; it is a data-dependent construction whose location is fixed by the fitted θMLE. The reported uncertainty estimates are then not independent epistemic statements about the data, but a consequence of combining the likelihood with a prior centered on the same data.

  2. renaming known result [Eq. (5) and Section 3.1, Figures 4-5; Appendix D, Table D.8]
    "π(t = 1|x,D) = 1/J Σ_{j=1}^J σ(⟨θ*_j, ψ_{θ_MLE}(x)⟩) ... 'when handling RA sentences, the Bayesian model exhibits greater uncertainty. Specifically... the mean activations for semantically plausible agents and patients decrease, along with an increase in the activation of syntactically indicated agents and patients.'"

    The predictive distribution is the average of sigmoid outputs over an ensemble with nonzero covariance. Averaging sigmoids over a spread of logits compresses extreme probabilities toward the middle: a high MLE activation (e.g., semantically plausible agent at ~0.9) falls, while a near-zero MLE activation (e.g., syntactically indicated agent) rises. The RA pattern described as 'greater uncertainty' is therefore a generic mathematical property of Eq. (5), not an empirical discovery about human-like processing. Appendix D confirms this: changing the arbitrary prior covariance to Pprior = 0.01I removes the effect (p > 0.001), showing that the central result is controlled by the injected prior variance rather than by data or by a validated posterior.

full rationale

The paper contains real statistical work, and its sampler is a legitimate method from prior literature. However, the central empirical claim is substantially produced by the construction itself. The prior is centered at the MLE trained on the same data D, so the resulting 'posterior' is not a Bayesian posterior under an independent pre-data belief; it is a data-dependent object whose mean is tied to the fitted solution. More importantly, the predictive distribution in Eq. (5) averages sigmoids over an ensemble with a hand-chosen covariance Pprior = I. This averaging necessarily softens extreme MLE outputs, which is exactly the RA pattern the paper highlights. Appendix D is decisive: with Pprior = 0.01I, the paired differences between Bayesian and ADAM activations are no longer significant, while larger priors produce highly significant differences. Thus the 'greater uncertainty' result is forced by the prior scale rather than emerging from an independently validated posterior or from human behavioral data. The paper presents this model-averaging smoothing as a step toward human-like comprehension, but no human data are collected; the qualitative link to Ferreira (2003) and Kuperberg et al. (2003) is borrowed rather than tested. This is a partial circularity: the method and its hyperparameters build in the softening that is then reported as evidence. It is not a full 8-10 because the sampler itself is a substantive numerical contribution and the direction of the reported shifts is not, strictly, identical to the prior mean; nevertheless, the main predictive claim reduces by construction to the combination of MLE-centering, arbitrary prior variance, and sigmoid ensembling.

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

The central result rests on the correctness of the cited sampler, the sufficiency of last-layer Bayesian inference, the interpretation of output activations as human-like uncertainty, and the adequacy of the synthetic corpus. The ad hoc prior centered at the MLE is the least standard ingredient and is treated as a free parameter.

free parameters (5)
  • Dropout rate rho = 0.3
    Hand-chosen in Section 3; modifies ensemble covariance in equation (14) and affects how strongly the sampler leaves the span of the initial ensemble.
  • Number of time steps tau = 17 discrete steps
    Stated in Section 3; the sampler is run for a finite horizon rather than until full convergence, so the posterior approximation depends on this choice.
  • Prior covariance scale Pprior = I (with 0.01 I and 5 I also tested)
    Chosen in Section 3; Appendix D shows the headline difference vanishes for 0.01 I, so this scale largely controls the reported effect.
  • Prior mean mprior = theta_MLE from phase-1 training
    Set equal to the MLE weight vector in Section 3; this uses the training data twice and anchors the posterior at the point estimate.
  • Ensemble size J = not stated
    The number of particles is never reported, although equation (5) and Algorithm 1 depend on it for the Monte Carlo approximation.
assumptions (5)
  • domain assumption The dropout deterministic sampler converges to samples from the posterior as tau tends to infinity.
    Invoked in Section 2.1.1; the convergence result is cited to Bhandari et al. (2024) and not re-derived.
  • domain assumption A last-layer Bayesian treatment with fixed MLE feature map captures the relevant model uncertainty.
    Section 2 states the first r-1 layers are trained by MLE and only output layer weights are inferred; uncertainty in features is assumed negligible.
  • domain assumption Model output activations and their spread can be interpreted as human-like uncertainty.
    Sections 1.1 and 4 link activation changes to human error rates and ERPs qualitatively, without measuring human participants.
  • ad hoc to paper A Gaussian prior centered at the MLE estimate is a valid prior for Bayesian inference on the same data.
    Section 3 sets mprior = theta_MLE; this uses the training data twice and is not a pre-data prior.
  • domain assumption The synthetic corpus of 10,000 sentences over 72 tokens is an adequate environment for studying sentence comprehension uncertainty.
    Section 2.3; generalization to natural language is assumed, not demonstrated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Ensemble Kalman filter for uncertainty in human language comprehension." pith.science (2026). https://pith.science/paper/SS5UVQHD

@misc{pith2026250502590,
  author       = {Pith},
  title        = {Pith review of: Ensemble Kalman filter for uncertainty in human language comprehension},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/SS5UVQHD}},
  note         = {Machine review of arXiv:2505.02590}
}
read the original abstract

Artificial neural networks (ANNs) are widely used in modeling sentence processing but often exhibit deterministic behavior, contrasting with human sentence comprehension, which manages uncertainty during ambiguous or unexpected inputs. This is exemplified by reversal anomalies-sentences with unexpected role reversals that challenge syntax and semantics-highlighting the limitations of traditional ANN models, such as the Sentence Gestalt (SG) Model. To address these limitations, we propose a Bayesian framework for sentence comprehension, applying an extension of the ensemble Kalman filter (EnKF) for Bayesian inference to quantify uncertainty. By framing language comprehension as a Bayesian inverse problem, this approach enhances the SG model's ability to reflect human sentence processing with respect to the representation of uncertainty. Numerical experiments and comparisons with maximum likelihood estimation (MLE) demonstrate that Bayesian methods improve uncertainty representation, enabling the model to better approximate human cognitive processing when dealing with linguistic ambiguities.

Figures

Figures reproduced from arXiv: 2505.02590 by the authors.

Figure 1
Figure 1. The Sentence Gestalt model data to learn the posterior distribution πpost. Using Bayes’ theorem, the posterior distribution is given by πpost(θ) ∝ exp(−l(θ)) πprior(θ), (1) where l(θ) denotes the negative log-likelihood function. Consider the data set D = {(ψ n , t n )} N n=1 , where t n are targets for the binary classification with input features ψ n ∈ R D for n = 1, ..., N. The ψ n ’s can either be the representa… view at source ↗
Figure 2
Figure 2. Last-layer Bayesian inference in SGM using ensemble transform methods [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Training & test loss and accuracy obtained by training SGM with ADAM optimizer. [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Mean activations (with standard deviations) of selected output units for the congruent form of test sentences [PITH_FULL_IMAGE:figures/full_fig_p012_4.png]
Figure 5
Figure 5. Figure 5: Mean activations (with standard deviations) of selected output units for the RA form of test sentences using [PITH_FULL_IMAGE:figures/full_fig_p013_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 15 canonical work pages

  1. [1]

    , author Pidstrigach, J

    author Bhandari, D. , author Pidstrigach, J. , author Reich, S. , year 2024 . title Affine invariant ensemble transform methods to improve predictive uncertainty in neural networks . journal Foundations of Data Science https://www.aimsciences.org/article/id/66e2a5bc2bcf34161cd7c93a, :10.3934/fods.2024040

    work page doi:10.3934/fods.2024040 2024

  2. [2]

    , author Nalisnick, E

    author Daxberger, E. , author Nalisnick, E. , author Allingham, J. , author Antorán, J. , author Hernández-Lobato, J.M. , year 2021 . title Bayesian deep learning via subnetwork inference , in: booktitle Proceedings of 38th International Conference on Machine Learning (ICML) , publisher PMLR . pp. pages 2510--2521 . https://proceedings.mlr.press/v139/daxb...

    work page 2021

  3. [3]

    , author Li, Q

    author Ding, Z. , author Li, Q. , year 2021 . title Ensemble kalman sampler: Mean-field limit and convergence analysis . journal SIAM Journal on Mathematical Analysis volume 53 , pages 1546--1578 . https://doi.org/10.1137/20M1339507, :10.1137/20M1339507

    work page doi:10.1137/20m1339507 2021

  4. [4]

    , year 2006

    author Evensen, G. , year 2006 . title Data Assimilation: The Ensemble Kalman Filter . publisher Springer-Verlag , address Berlin, Heidelberg

    work page 2006

  5. [5]

    , year 2003

    author Ferreira, F. , year 2003 . title The misinterpretation of noncanonical sentences . journal Cognitive Psychology volume 47 , pages 164--203 . https://doi.org/10.1016/S0010-0285(03)00005-7, :10.1016/S0010-0285(03)00005-7

    doi:10.1016/s0010-0285(03)00005-7 2003

  6. [6]

    , year 2016

    author Gal, Y. , year 2016 . title Uncertainty in Deep Learning . Ph.D. thesis. University of Cambridge

    work page 2016

  7. [7]

    , author Ghahramani, Z

    author Gal, Y. , author Ghahramani, Z. , year 2016 . title Dropout as a bayesian approximation: Representing model uncertainty in deep learning , in: editor Balcan, M.F. , editor Weinberger, K.Q. (Eds.), booktitle Proceedings of The 33rd International Conference on Machine Learning , publisher PMLR , address New York, New York, USA . pp. pages 1050--1059 ...

    work page 2016

  8. [8]

    , author Pleiss, G

    author Guo, C. , author Pleiss, G. , author Sun, Y. , author Weinberger, K.Q. , year 2017 . title On calibration of modern neural networks , in: editor Precup, D. , editor Teh, Y.W. (Eds.), booktitle Proceedings of the 34th International Conference on Machine Learning , publisher PMLR . pp. pages 1321--1330 . https://proceedings.mlr.press/v70/guo17a.html

    work page 2017

Show all 21 references
  1. [9]

    , author Lucka, F

    author Haber, E. , author Lucka, F. , author Ruthotto, L. , year 2018 . title Never look back - a modified enkf method and its application to the training of neural networks without back propagation . http://arxiv.org/abs/1805.08034 arXiv:1805.08034

    2018 arXiv

  2. [10]

    , author Andriushchenko, M

    author Hein, M. , author Andriushchenko, M. , author Bitterwolf, J. , year 2019 . title Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem , in: booktitle Proceedings of the IEEE/CVF Conference on Computer Vision...

    2019

  3. [11]

    , author Stuart, A.M

    author Kovachki, N.B. , author Stuart, A.M. , year 2019 . title Ensemble kalman inversion: a derivative-free technique for machine learning tasks . journal Inverse Problems volume 35 , pages 095005 . https://dx.doi.org/10.1088/1361-6420/ab1c3a, :10.1088/1361-6420/ab1c3a

    2019 doi

  4. [12]

    , author Hein, M

    author Kristiadi, A. , author Hein, M. , author Hennig, P. , year 2020 . title Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks , in: booktitle Proceedings of the 37th International Conference on Machine Learning

    2020

  5. [13]

    , author Sitnikova, T

    author Kuperberg, G.R. , author Sitnikova, T. , author Caplan, D. , author Holcomb, P.J. , year 2003 . title Electrophysiological distinctions in processing conceptual relationships within simple sentences . journal Cognitive Brain Research volume 17 , pages 117--129 . :10.101...

    2003 doi

  6. [14]

    , author Federmeier, K.D

    author Kutas, M. , author Federmeier, K.D. , year 2011 . title Thirty years and counting: finding meaning in the N400 component of the event-related brain potential ( ERP ) . journal Annu. Rev. Psychol. volume 62 , pages 621--647

    2011

  7. [15]

    , author St

    author McClelland, J.L. , author St. John, M. , author Taraban, R. , year 1989 . title Sentence comprehension: A parallel distributed processing approach . journal Language and Cognitive Processes volume 4 , pages SI287--SI335 . https://doi.org/10.1080/01690968908406371, :10.1...

    1989 doi

  8. [16]

    , year 2011

    author Neal, R.M. , year 2011 . title MCMC Using Hamiltonian Dynamics . publisher Chapman and Hall/CRC . p. pages 113–162 . http://dx.doi.org/10.1201/b10905-6, :10.1201/b10905-6

    2011 doi

  9. [17]

    , year 2012

    author Neal, R.M. , year 2012 . title Bayesian learning for neural networks . volume volume 118 . publisher Springer Science & Business Media

    2012

  10. [18]

    , author Reich, S

    author Pidstrigach, J. , author Reich, S. , year 2022 . title Affine-invariant ensemble transform methods for logistic regression . journal Foundations of Computational Mathematics , pages 1--34

    2022

  11. [19]

    , author Hansen, S

    author Rabovsky, M. , author Hansen, S. , author Mcclelland, J. , year 2018 . title Modelling the N400 brain potential as change in a probabilistic representation of meaning . journal Nature Human Behaviour :10.1038/s41562-018-0406-4

    2018 doi

  12. [20]

    , author Hinton, G

    author Srivastava, N. , author Hinton, G. , author Krizhevsky, A. , author Sutskever, I. , author Salakhutdinov, R. , year 2014 . title Dropout: A simple way to prevent neural networks from overfitting . journal Journal of Machine Learning Research volume 15 , pages 1929--1958...

    2014

  13. [21]

    John , M.F

    author St. John , M.F. , author McClelland, J.L. , year 1990 . title Learning and applying contextual constraints in sentence comprehension . journal Artificial Intelligence volume 46 , pages 217--257 . https://www.sciencedirect.com/science/article/pii/000437029090008N, :https...

    1990

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.