Pith. sign in

REVIEW 1 major objections 3 minor 75 references

Fast Multi-Group Gaussian Process Factor Models

T0 review · 1 major / 3 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read This paper claims that multi-group Gaussian process factor models can be fit in time linear in trial length and group count using inducing variables or a frequency-domain approximation, with orders-of-magnitude speedups and negligible…

desk verdict Frequency-domain acceleration is solid and useful; the abstract overstates the inducing-variable variant. read the letter →

arxiv 2412.16773 v1 pith:ENVJ7PWJ submitted 2024-12-21 stat.ML cs.LGeess.SPq-bio.NC

Evren Gokcen , Anna I. Jasper , Adam Kohn , Christian K. Machens , Byron M. Yu This is my paper · ORCID
classification stat.MLcs.LGeess.SPq-bio.NC MSC 60G1562F1562H25
keywords Gaussianprocessfactormodelsmulti-groupdimensionalityreductioninducingvariablesfrequencydomainWhittlelikelihoodlinear-timescalingneuralpopulationrecordingsautomaticrelevancedetermination
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to remove the computational bottleneck in a class of dimensionality-reduction models used to study how multiple neural populations interact over time. The baseline method, called mDLAG-time, must invert covariance matrices that grow cubically in the number of time points per trial and in the number of recorded groups, which makes large multi-area recordings impractical. The paper proposes two approximations: one that passes latent structure through a small set of inducing variables, and one that formulates the entire generative model in the frequency domain, where the GP covariance becomes diagonal. Both achieve linear scaling in trial length and group number. On simulated data and on Neuropixels recordings of hundreds of neurons across three visual areas, the frequency-domain method delivers the largest speedups (up to about 25x overall on real data) while matching the predictive performance of the exact time-domain fit.

What carries the argument

The load-bearing object is the frequency-domain (Whittle) approximation to the Gaussian process state model: for a stationary GP, the unitary DFT of the covariance matrix converges to a diagonal matrix whose entries are the power spectral density, so the multi-group covariance is replaced by a diagonal PSD matrix. Because the time-delay terms factor as a product of per-group phase-shift matrices, the posterior covariance at each frequency is a p by p matrix inverse rather than a pMT by pMT inverse, which is what turns O($M^{3}$ $T^{3}$) into O(M T) per iteration. The inducing-variable variant instead replaces the MT-dimensional latent trajectory with a small number of inducing points per latent, giving O(M T $T_ind^{2}$) operations and preserving the ARD-based group-sparsity structure of the original model.

What would settle it

In the paper's own trial-length sweep, mDLAG-frequency's leave-group-out R2 is already about 1 percent below mDLAG-time at T=10 (Fig. 3A inset); a dataset with short trials and fast latent timescales where this gap reaches a practically meaningful level at T at least 100, or where estimated GP timescales stay biased by more than about 10 percent even at long T, would falsify the claim that the frequency-domain fit preserves statistical performance.

Watch

Extended reading notes

Core claim

The central claim is that the cubic runtime scaling that has limited multi-group Gaussian process factor models can be broken: mDLAG-inducing, built on variational inducing variables, and mDLAG-frequency, a fully frequency-domain generative model, both reduce the cost of the three bottleneck stages of fitting (latent posterior update, GP-parameter gradients, and ELBO evaluation) to O($p^{3}$ T + $p^{2}$ M T) and O(M T) per iteration respectively, i.e., linear in the number of time points per trial T and the number of groups M. The frequency-domain construction is the more complete of the two: by taking the GP covariance to be diagonal in the Fourier basis (the Whittle / circulant approximation), the posterior over latents factorizes across frequencies, and the time-delay parameters enter only as diagonal phase-shift matrices, so each frequency update costs O($p^{3}$) plus O($p^{2}$ M). Empirically the paper shows that mDLAG-frequency matches mDLAG-time's leave-group-out R2 on long simulated trials and on three-area Neuropixels data while running about 25x faster end-to-end, and it characterizes the cost of the approximation: at short trial lengths, GP timescales and delays are underestimated and latent dimensionality can be overestimated, biases that shrink as T grows and that tapering largely removes.

Load-bearing premise

The central speedups rest on two approximations whose error grows as trials get shorter: the frequency-domain method assumes the GP covariance is diagonal (circulant in time, exact only as T grows), and the inducing-variable method assumes the number of inducing points can stay fixed without aliasing fast latent timescales.

Editorial extensions

If this is right

  • Multi-group GP factor models can be applied to trials with hundreds of time points and to recordings spanning many brain areas, where the baseline method would require days of compute.
  • The frequency-domain fit can replace the time-domain fit in exploratory analyses: on the datasets tested, leave-group-out R2 is statistically indistinguishable while runtime drops by orders of magnitude.
  • The same two approximation strategies carry over to the broader GP factor model family (GPFA, DLAG, group factor analysis), since mDLAG subsumes them as special cases.
  • Users of these methods must be aware of the frequency-domain biases at short trial lengths, and can mitigate them with tapering or by fine-tuning with the time-domain model.

Reading between the lines

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

  • The frequency-domain method's scaling advantage grows exactly where its approximation is most accurate, so the practical sweet spot is long trials; a hybrid protocol of fitting with mDLAG-frequency for fast exploration and then fine-tuning the retained model with mDLAG-time could capture the best of both.
  • Because the same diagonal-covariance trick is what makes the approach linear in the number of groups, the method should extend to non-Gaussian observation models (e.g., Poisson spike counts) via a frequency-domain quasi-likelihood, though the paper only demonstrates the Gaussian case.
  • The bias characterization implies that reported GP timescales and delays from frequency-domain fits should be interpreted with a finite-T correction, and that cross-validated dimensionality selection via ARD may overstate the number of shared latents on short trials.
  • The inducing-variable approach's need to keep the number of inducing points above the Nyquist rate couples its runtime to the fastest latent timescale, so on data with fast timescales the frequency-domain method is likely to remain the better default, a hypothesis testable by comparing the two on a dataset with known fast and slow latents.
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

1 major / 3 minor

Summary. The paper proposes two approximate inference algorithms for multi-group Gaussian process factor models of the type used in multi-area neural recordings: mDLAG-inducing, which introduces a shared inducing variable per latent, and mDLAG-frequency, which formulates the model directly in the Fourier domain under a diagonal-spectrum (circulant) approximation to the GP covariance. The authors derive variational updates and gradient-based GP parameter updates for both methods, state per-iteration complexity claims of linear scaling in trial length T and group number M, and evaluate the methods on simulated data and Neuropixels recordings from three visual cortical areas. They also characterize finite-T biases in mDLAG-frequency's GP timescale, time delay, and dimensionality estimates and show that tapering mitigates these biases. The central claims are that both accelerated methods preserve the statistical performance of the time-domain baseline while achieving orders-of-magnitude speed-ups, with mDLAG-frequency offering the best overall trade-off.

Significance. If the speed and accuracy claims hold as stated, the paper would make a practically important class of dimensionality-reduction methods feasible for large multi-population recordings. The variational derivations in Appendices B and C are detailed, the scaling analyses are internally consistent under their stated assumptions, and the frequency-domain method is supported by a thorough bias analysis and a reproducible-looking evaluation protocol. However, the paper's abstract and introduction overstate the linear-scaling evidence for mDLAG-inducing, and the frequency-domain method's finite-T parameter biases are substantial enough that the blanket claim of 'minimal impact on statistical performance' needs sharp scoping. With those qualifications addressed, the contribution is significant for the computational neuroscience and GP latent-variable-model communities.

major comments (1)
  1. [Section 5, first paragraph] The Discussion repeats the overstatement: 'We demonstrated that mDLAG-inducing and mDLAG-frequency can achieve linear scaling in both the number of time points per trial T and the number of groups M'. This is true for mDLAG-inducing only conditionally on Tind being fixed, and the subsequent paragraphs correctly describe the practical trade-off. The summary paragraph should be amended so that the main claim matches the evidence presented in Sections 4.2 and 4.4.
minor comments (3)
  1. [Section 4.4] The text says mDLAG-frequency 'performed statistically as well as mDLAG-time' based on Fig. 5B, but no paired statistical test is reported for that comparison; adding the sign-test p-value or a confidence interval would make the claim precise.
  2. [Section 4.1; Fig. 2C] The latent time-course R2 values are computed after re-inferring all latents with mDLAG-time inference (equation 3.13), even for models fit by mDLAG-inducing and mDLAG-frequency. This is a reasonable way to compare parameter estimates, but it does not evaluate the latent estimates users would obtain from each method's native inference. A sentence stating this explicitly would prevent misinterpretation.
  3. [Reproducibility section] The implementations of mDLAG-inducing and mDLAG-frequency are listed as 'will be made available upon publication'; making the code available at submission would strengthen the reproducibility of the empirical claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the scaling derivations are self-contained, and the empirical claims are supported by held-out test data and real Neuropixels recordings, with the paper's own caveats explicitly disclosed.

full rationale

The core derivation steps are internally consistent and do not reduce to their inputs by construction. The mDLAG-inducing complexity bound in Section 3.2.2 and Appendix B is explicitly O(p^3 M T Tind^2) plus O(p^3 Tind^3), which is linear in T only for fixed Tind; the paper itself states this conditional structure and then reports that Tind must be increased with T to avoid aliasing, producing superlinear practical scaling (Section 4.2, Supplementary Fig. 2) and a runtime/accuracy trade-off on real data (Section 4.4, Supplementary Fig. 4). This is an honest limitation of a conditional claim, not a circular reduction. The mDLAG-frequency update (Eq. 3.33) is genuinely diagonal in frequency, giving the stated O(p^3 T + p^2 M T) cost, and the diagonal-spectrum approximation is justified by external classical spectral results (Kolmogorov 1941; Priestley 1981; Whittle 1951), not by a self-citation chain. Statistical evaluation uses held-out test trials, including leave-group-out R2 on Neuropixels recordings (Section 4.4, Fig. 5), so the main performance claims are not fitted inputs renamed as predictions. The demo simulation (Section 4.1) is generated from the time-domain mDLAG model, and the larger simulations are generated from the mDLAG-frequency formulation as explicitly disclosed in Section 6.1, with edge effects mitigated by generating longer trials and retaining only the middle third; this is a self-consistency check on the method's own assumptions rather than a derivation-level circularity, and it is balanced by the real-data results. Baseline mDLAG is cited from prior published work (Gokcen et al., 2023) rather than being derived from unsupported self-citations. No uniqueness theorem, ansatz, or known result is smuggled in via author self-citation. Overall, the paper's approximations and their limitations are stated transparently, and the central claims have independent content.

Assumptions & free parameters 4 free parameters · 6 assumptions · 2 invented entities

The central complexity claims rest on standard linear algebra and Fourier analysis; the empirical accuracy claims rest on the mDLAG generative model, the diagonal-spectrum approximation, and dataset and hyperparameter choices. The most consequential uncharged inputs are the finite-T circulant approximation and the hand-selected Tind, both of which the paper itself shows can introduce bias.

free parameters (4)
  • Tind (number of inducing points) = 20, 32, 50, 13 depending on dataset
    Hand-chosen hyperparameter for mDLAG-inducing; controls accuracy versus runtime and must be large enough to avoid aliasing. The paper adjusts it per experiment, so the linear-in-T claim is conditional on Tind being held fixed.
  • Latent significance threshold (nu_min) = 0.02
    Latents explaining less than 2% of shared variance in a group are discarded (Methods 6.3). Changing this threshold changes reported dimensionality and model-selection conclusions.
  • Dmax delay bound = Half the trial length
    Time delays are constrained to [-Dmax, Dmax] via a tanh transform (Appendix C). This is a hand-chosen identifiability/plausibility bound that affects delay estimates.
  • Convergence tolerance = 1e-8
    Fitting stops when the lower bound improves by less than 1e-8. Runtime-to-convergence comparisons depend on this tolerance, and the three methods optimize different lower bounds, making iteration counts not directly commensurable.
assumptions (6)
  • standard math The DFT of a stationary GP has a covariance matrix that approaches a diagonal matrix as T grows (Whittle approximation).
    Invoked in Section 3.3.1 to justify the diagonal PSD matrix in Eq. 3.25; the paper explicitly notes it is an approximation for finite T.
  • domain assumption The mDLAG generative model: latents are shared across groups with stationary squared-exponential GP covariance and additive time delays, with ARD priors on loadings.
    This is the baseline model from Gokcen et al. (2023), and all three fitting methods inherit it; if the true neural interactions are not well described by this model, the methods' usefulness is limited.
  • domain assumption Mean-field variational factorization (Eq. 3.11, 3.21, 3.31) with noninformative hyperparameters fixed to 1e-12.
    This factorization enables closed-form updates but is a standard approximation to the posterior; the paper does not quantify its error.
  • ad hoc to paper For mDLAG-frequency, replacing the time-domain covariance with a circulant matrix (diagonal in the Fourier domain) yields a usable generative model and inference procedure.
    This is the core approximation (Eq. 3.25), known to introduce edge effects and parameter biases at finite T; the paper characterizes and mitigates but does not eliminate it.
  • ad hoc to paper For mDLAG-inducing, a single common inducing variable per latent, sampled at Tind points, can represent the delayed latents across all M groups.
    This is the key structural simplification (Eqs. 3.15-3.20); its accuracy depends on Tind being large enough relative to the latent timescale, which is not known a priori.
  • domain assumption Tind must be chosen above the approximate Nyquist rate of the latent signals to avoid aliasing.
    The paper uses 12-13 Hz as the approximate Nyquist rate for 100 ms squared-exponential GPs (Sec. 4.2); this is an empirical rule, not a theorem, and it forces Tind to grow with trial length in practice.
invented entities (2)
  • Shared inducing variable w_{n,j,:}
    purpose: A low-dimensional temporal skeleton common to all groups, used by mDLAG-inducing to reduce covariance size and achieve linear scaling in M and T.
    Standard sparse-GP auxiliary variables, not physically observable; no falsifiable prediction outside the model is provided.
  • Frequency-domain latent ex_{n,j,:} with diagonal PSD
    purpose: A Fourier-domain representation of the latent GP that makes the prior covariance diagonal, enabling O(MT) inference for mDLAG-frequency.
    This is a mathematical reparameterization of the same latent process, not a new physical entity; it lacks independent observable evidence beyond the model itself.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fast Multi-Group Gaussian Process Factor Models." pith.science (2026). https://pith.science/paper/ENVJ7PWJ

@misc{pith2026241216773,
  author       = {Pith},
  title        = {Pith review of: Fast Multi-Group Gaussian Process Factor Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ENVJ7PWJ}},
  note         = {Machine review of arXiv:2412.16773}
}
read the original abstract

Gaussian processes are now commonly used in dimensionality reduction approaches tailored to neuroscience, especially to describe changes in high-dimensional neural activity over time. As recording capabilities expand to include neuronal populations across multiple brain areas, cortical layers, and cell types, interest in extending Gaussian process factor models to characterize multi-population interactions has grown. However, the cubic runtime scaling of current methods with the length of experimental trials and the number of recorded populations (groups) precludes their application to large-scale multi-population recordings. Here, we improve this scaling from cubic to linear in both trial length and group number. We present two approximate approaches to fitting multi-group Gaussian process factor models based on (1) inducing variables and (2) the frequency domain. Empirically, both methods achieved orders of magnitude speed-up with minimal impact on statistical performance, in simulation and on neural recordings of hundreds of neurons across three brain areas. The frequency domain approach, in particular, consistently provided the greatest runtime benefits with the fewest trade-offs in statistical performance. We further characterize the estimation biases introduced by the frequency domain approach and demonstrate effective strategies to mitigate them. This work enables a powerful class of analysis techniques to keep pace with the growing scale of multi-population recordings, opening new avenues for exploring brain function.

Figures

Figures reproduced from arXiv: 2412.16773 by the authors.

Figure 1
Figure 1. Summary of generative models and covariance structures. (A) Delayed latents across [PITH_FULL_IMAGE:figures/full_fig_p008_1.png] view at source ↗
Figure 2
Figure 2. Demonstrative simulation. (A) Loading matrix estimates. From left to right: Ground [PITH_FULL_IMAGE:figures/full_fig_p022_2.png] view at source ↗
Figure 3
Figure 3. Scaling of performance and runtime with number of time points per trial. (A) Leave [PITH_FULL_IMAGE:figures/full_fig_p025_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Scaling of performance and runtime with number of groups. (A) Leave-unit-out [PITH_FULL_IMAGE:figures/full_fig_p027_4.png]
Figure 5
Figure 5. Figure 5: Performance and runtime on Neuropixels recordings from macaque visual cortex. Here [PITH_FULL_IMAGE:figures/full_fig_p029_5.png]
Figure 6
Figure 6. Figure 6: The circulant approximation induced by the frequency domain approach and its effects [PITH_FULL_IMAGE:figures/full_fig_p031_6.png]
Figure 7
Figure 7. Figure 7: Bias in Gaussian process (GP) parameter estimation with number of time points per [PITH_FULL_IMAGE:figures/full_fig_p033_7.png]
Figure 8
Figure 8. Figure 8: Bias in estimation of latent dimensionality with number of time points per trial. (A) [PITH_FULL_IMAGE:figures/full_fig_p034_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

75 extracted references · 59 canonical work pages

  1. [1]

    , Li, J M

    ahrens_brain-wide_2012 APACrefauthors Ahrens, M B. , Li, J M. , Orger, M B. , Robson, D N. , Schier, A F. , Engert, F. \ Portugues, R. APACrefauthors \ 2012 . Brain-wide neuronal dynamics during motor adaptation in zebrafish Brain-wide neuronal dynamics during motor adaptation in zebrafish . Nature 485 7399 471--477 . APACrefDOI doi:10.1038/nature11057 APACrefDOI

    doi:10.1038/nature11057 2012

  2. [2]

    , Luengo, D

    alvarez_efficient_2010 APACrefauthors Álvarez, M. , Luengo, D. , Titsias, M. \ Lawrence, N D. APACrefauthors \ 2010 . Efficient Multioutput Gaussian Processes through Variational Inducing Kernels Efficient Multioutput Gaussian Processes through Variational Inducing Kernels . Proceedings of the Thirteenth International Conference on Artificial Intelligence...

    work page 2010

  3. [3]

    , Rosasco, L

    alvarez_kernels_2012 APACrefauthors Álvarez, M. , Rosasco, L. \ Lawrence, N D. APACrefauthors \ 2012 . Kernels for Vector - Valued Functions : A Review Kernels for Vector - Valued Functions : A Review . Foundations and Trends in Machine Learning 4 3 195--266 . APACrefDOI doi:10.1561/2200000036 APACrefDOI

    doi:10.1561/2200000036 2012

  4. [4]

    \ Pillow, J W

    aoi_scalable_2017 APACrefauthors Aoi, M C. \ Pillow, J W. APACrefauthors \ 2017 . Scalable Bayesian inference for high-dimensional neural receptive fields Scalable Bayesian inference for high-dimensional neural receptive fields . bioRxiv . APACrefDOI doi:10.1101/212217 APACrefDOI

    work page doi:10.1101/212217 2017

  5. [5]

    , Noel, J P G

    balzani_probabilistic_2023 APACrefauthors Balzani, E. , Noel, J P G. , Herrero-Vidal, P. , Angelaki, D E. \ Savin, C. APACrefauthors \ 2023 . A probabilistic framework for task-aligned intra- and inter-area neural manifold estimation A probabilistic framework for task-aligned intra- and inter-area neural manifold estimation . The Eleventh International Co...

    work page 2023

  6. [6]

    APACrefauthors \ 1999

    bishop_variational_1999 APACrefauthors Bishop, C. APACrefauthors \ 1999 . Variational principal components Variational principal components . 9th International Conference on Artificial Neural Networks 509--514 . APACrefDOI doi:10.1049/cp:19991160 APACrefDOI

    work page doi:10.1049/cp:19991160 1999

  7. [7]

    , Liu, Z

    bong_latent_2020 APACrefauthors Bong, H. , Liu, Z. , Ren, Z. , Smith, M. , Ventura, V. \ Robert, K E. APACrefauthors \ 2020 . Latent Dynamic Factor Analysis of High - Dimensional Neural Recordings Latent Dynamic Factor Analysis of High - Dimensional Neural Recordings . Advances in Neural Information Processing Systems 33 16446--16456

    work page 2020

  8. [8]

    , Snyder, A C

    cowley_slow_2020 APACrefauthors Cowley, B R. , Snyder, A C. , Acar, K. , Williamson, R C. , Yu, B M. \ Smith, M A. APACrefauthors \ 2020 . Slow Drift of Neural Activity as a Signature of Impulsivity in Macaque Visual and Prefrontal Cortex Slow Drift of Neural Activity as a Signature of Impulsivity in Macaque Visual and Prefrontal Cortex . Neuron 108 3 551...

    work page doi:10.1016/j.neuron.2020.07.021 2020

Show all 75 references
  1. [9]

    , Shenoy, K V

    cunningham_fast_2008 APACrefauthors Cunningham, J P. , Shenoy, K V. \ Sahani, M. APACrefauthors \ 2008 . Fast Gaussian process methods for point process intensity estimation Fast Gaussian process methods for point process intensity estimation . Proceedings of the 25th internat...

    2008

  2. [10]

    \ Yu, B M

    cunningham_dimensionality_2014 APACrefauthors Cunningham, J P. \ Yu, B M. APACrefauthors \ 2014 . Dimensionality reduction for large-scale neural recordings Dimensionality reduction for large-scale neural recordings . Nature Neuroscience 17 11 1500--1509 . APACrefDOI doi:10.10...

    2014 doi

  3. [11]

    \ Newsam, G N

    dietrich_fast_1997 APACrefauthors Dietrich, C R. \ Newsam, G N. APACrefauthors \ 1997 . Fast and Exact Simulation of Stationary Gaussian Processes through Circulant Embedding of the Covariance Matrix Fast and Exact Simulation of Stationary Gaussian Processes through Circulant ...

    1997 doi

  4. [12]

    , Zhao, Y

    dowling_linear_2023 APACrefauthors Dowling, M. , Zhao, Y. \ Park, I M. APACrefauthors \ 2023 . Linear Time GPs for Inferring Latent Trajectories from Neural Spike Trains Linear Time GPs for Inferring Latent Trajectories from Neural Spike Trains . Proceedings of the 40th Intern...

    2023

  5. [13]

    , Ruda, K M

    duncker_scalable_2023 APACrefauthors Duncker, L. , Ruda, K M. , Field, G D. \ Pillow, J W. APACrefauthors \ 2023 . Scalable Variational Inference for Low - Rank Spatiotemporal Receptive Fields Scalable Variational Inference for Low - Rank Spatiotemporal Receptive Fields . Neur...

    2023 doi

  6. [14]

    \ Sahani, M

    duncker_temporal_2018 APACrefauthors Duncker, L. \ Sahani, M. APACrefauthors \ 2018 . Temporal alignment and latent Gaussian process factor inference in population spike trains Temporal alignment and latent Gaussian process factor inference in population spike trains . Advance...

    2018

  7. [15]

    , Agol, E

    foreman-mackey_fast_2017 APACrefauthors Foreman-Mackey, D. , Agol, E. , Ambikasaran, S. \ Angus, R. APACrefauthors \ 2017 . Fast and Scalable Gaussian Process Modeling with Applications to Astronomical Time Series Fast and Scalable Gaussian Process Modeling with Applications t...

    2017 doi

  8. [16]

    , Ulrich, K R

    gallagher_cross-spectral_2017 APACrefauthors Gallagher, N. , Ulrich, K R. , Talbot, A. , Dzirasa, K. , Carin, L. \ Carlson, D E. APACrefauthors \ 2017 . Cross- Spectral Factor Analysis Cross- Spectral Factor Analysis . Advances in Neural Information Processing Systems 30 6842--6852

    2017

  9. [17]

    , Whiteway, M

    glaser_recurrent_2020 APACrefauthors Glaser, J. , Whiteway, M. , Cunningham, J P. , Paninski, L. \ Linderman, S. APACrefauthors \ 2020 . Recurrent Switching Dynamical Systems Models for Multiple Interacting Neural Populations Recurrent Switching Dynamical Systems Models for Mu...

    2020

  10. [18]

    APACrefauthors \ 2023

    gokcen_mdlag_2023 APACrefauthors Gokcen, E. APACrefauthors \ 2023 . mDLAG (v1.0.0) mDLAG (v1.0.0) . Zenodo DOI: 10.5281/zenodo.10048162

    2023 doi

  11. [19]

    , Jasper, A I

    gokcen_disentangling_2022 APACrefauthors Gokcen, E. , Jasper, A I. , Semedo, J D. , Zandvakili, A. , Kohn, A. , Machens, C K. \ Yu, B M. APACrefauthors \ 2022 . Disentangling the flow of signals between populations of neurons Disentangling the flow of signals between populatio...

    2022 doi

  12. [20]

    , Jasper, A I

    gokcen_uncovering_2023 APACrefauthors Gokcen, E. , Jasper, A I. , Xu, A. , Kohn, A. , Machens, C K. \ Yu, B M. APACrefauthors \ 2023 . Uncovering motifs of concurrent signaling across multiple neuronal populations Uncovering motifs of concurrent signaling across multiple neuro...

    2023

  13. [21]

    \ Van Loan, C F

    golub_matrix_2013 APACrefauthors Golub, G H. \ Van Loan, C F. APACrefauthors \ 2013 . Matrix computations Matrix computations \ ( 4 \ ). The Johns Hopkins University Press

    2013

  14. [22]

    , Sikandar, U B

    gondur_multi-modal_2024 APACrefauthors Gondur, R. , Sikandar, U B. , Schaffer, E. , Aoi, M C. \ Keeley, S L. APACrefauthors \ 2024 . Multi-modal Gaussian Process Variational Autoencoders for Neural and Behavioral Data Multi-modal Gaussian Process Variational Autoencoders for N...

    2024

  15. [23]

    , Sykulski, A M

    guillaumin_debiased_2022 APACrefauthors Guillaumin, A P. , Sykulski, A M. , Olhede, S C. \ Simons, F J. APACrefauthors \ 2022 . The Debiased Spatial Whittle likelihood The Debiased Spatial Whittle likelihood . Journal of the Royal Statistical Society. Series B, Statistical Met...

    2022 doi

  16. [24]

    \ Särkkä, S

    hartikainen_kalman_2010 APACrefauthors Hartikainen, J. \ Särkkä, S. APACrefauthors \ 2010 . Kalman filtering and smoothing solutions to temporal Gaussian process regression models Kalman filtering and smoothing solutions to temporal Gaussian process regression models . 2010 IE...

    2010

  17. [25]

    , Durrande, N

    hensman_variational_2018 APACrefauthors Hensman, J. , Durrande, N. \ Solin, A. APACrefauthors \ 2018 . Variational Fourier Features for Gaussian Processes Variational Fourier Features for Gaussian Processes . Journal of Machine Learning Research 18 151 1--52

    2018

  18. [26]

    , Fusi, N

    hensman_gaussian_2013 APACrefauthors Hensman, J. , Fusi, N. \ Lawrence, N D. APACrefauthors \ 2013 . Gaussian processes for Big data Gaussian processes for Big data . Proceedings of the Twenty-Ninth Conference on Uncertainty in Artificial Intelligence 282--290

    2013

  19. [27]

    , Kao, T C

    jensen_scalable_2021 APACrefauthors Jensen, K. , Kao, T C. , Stone, J. \ Hennequin, G. APACrefauthors \ 2021 . Scalable Bayesian GPFA with automatic relevance determination and discrete noise models Scalable Bayesian GPFA with automatic relevance determination and discrete noi...

    2021

  20. [28]

    \ Bucy, R S

    kalman_new_1961 APACrefauthors Kalman, R E. \ Bucy, R S. APACrefauthors \ 1961 . New Results in Linear Filtering and Prediction Theory New Results in Linear Filtering and Prediction Theory . Transactions of the ASME, Journal of Basic Engineering 83 1 95--108 . APACrefDOI doi:1...

    1961 doi

  21. [29]

    \ Druckmann, S

    kang_approaches_2020 APACrefauthors Kang, B. \ Druckmann, S. APACrefauthors \ 2020 . Approaches to inferring multi-regional interactions from simultaneous population recordings Approaches to inferring multi-regional interactions from simultaneous population recordings . Curren...

    2020 doi

  22. [30]

    , Zoltowski, D M

    karniol-tambour_modeling_2024 APACrefauthors Karniol-Tambour, O. , Zoltowski, D M. , Diamanti, E M. , Pinto, L. , Brody, C D. , Tank, D W. \ Pillow, J W. APACrefauthors \ 2024 . Modeling state-dependent communication between brain regions with switching nonlinear dynamical sys...

    2024

  23. [31]

    , Aoi, M

    keeley_identifying_2020 APACrefauthors Keeley, S. , Aoi, M. , Yu, Y. , Smith, S. \ Pillow, J W. APACrefauthors \ 2020 . Identifying signal and noise structure in neural population activity with Gaussian process factor models Identifying signal and noise structure in neural pop...

    2020

  24. [32]

    , Zoltowski, D M

    keeley_modeling_2020 APACrefauthors Keeley, S. , Zoltowski, D M. , Aoi, M C. \ Pillow, J W. APACrefauthors \ 2020 . Modeling statistical dependencies in multi-region spike train data Modeling statistical dependencies in multi-region spike train data . Current Opinion in Neurob...

    2020 doi

  25. [33]

    , Zoltowski, D M

    keeley_efficient_2020 APACrefauthors Keeley, S. , Zoltowski, D M. , Yu, Y. , Yates, J L. , Smith, S L. \ Pillow, J W. APACrefauthors \ 2020 . Efficient non-conjugate Gaussian process factor models for spike count data using polynomial approximations Efficient non-conjugate Gau...

    2020

  26. [34]

    , Virtanen, S

    klami_group_2015 APACrefauthors Klami, A. , Virtanen, S. , Leppäaho, E. \ Kaski, S. APACrefauthors \ 2015 . Group Factor Analysis Group Factor Analysis . IEEE Transactions on Neural Networks and Learning Systems 26 9 2136--2147 . APACrefDOI doi:10.1109/TNNLS.2014.2376974 APACrefDOI

    2015

  27. [35]

    APACrefauthors \ 1941

    kolmogorov_stationary_1941 APACrefauthors Kolmogorov, A. APACrefauthors \ 1941 . Stationary sequences in Hilbert space Stationary sequences in Hilbert space . Moscow University Mathematics Bulletin 2 6

    1941

  28. [36]

    , Sadtler, P T

    lakshmanan_extracting_2015 APACrefauthors Lakshmanan, K C. , Sadtler, P T. , Tyler-Kabara, E C. , Batista, A P. \ Yu, B M. APACrefauthors \ 2015 . Extracting Low - Dimensional Latent Structure from Time Series in the Presence of Delays Extracting Low - Dimensional Latent Struc...

    2015 doi

  29. [37]

    li_multi-region_2024 APACrefauthors Li, W. , Li, C. , Wang, Y. \ Wu, A. APACrefauthors \ 2024 . Multi- Region Markovian Gaussian Process : An Efficient Method to Discover Directional Communications Across Multiple Brain Regions Multi- Region Markovian Gaussian Process : An Eff...

    2024

  30. [38]

    , Wang, Y

    li_markovian_2024 APACrefauthors Li, W. , Wang, Y. , Li, C. \ Wu, A. APACrefauthors \ 2024 . Markovian Gaussian Process : A Universal State - Space Representation for Stationary Temporal Gaussian Process Markovian Gaussian Process : A Universal State - Space Representation for...

    2024 doi arXiv

  31. [39]

    , Park, J

    lim_multi-output_2021 APACrefauthors Lim, J. , Park, J. , Nah, S. \ Choi, J. APACrefauthors \ 2021 . Multi-output Infinite Horizon Gaussian Processes Multi-output Infinite Horizon Gaussian Processes . 2021 IEEE International Conference on Robotics and Automation (ICRA) 1542--1...

    2021

  32. [40]

    , Blei, D

    loper_general_2021 APACrefauthors Loper, J. , Blei, D. , Cunningham, J P. \ Paninski, L. APACrefauthors \ 2021 . A general linear-time inference method for Gaussian Processes on one dimension A general linear-time inference method for Gaussian Processes on one dimension . Jour...

    2021

  33. [41]

    , Kauvar, I V

    machado_multiregion_2022 APACrefauthors Machado, T A. , Kauvar, I V. \ Deisseroth, K. APACrefauthors \ 2022 . Multiregion neuronal activity: the forest and the trees Multiregion neuronal activity: the forest and the trees . Nature Reviews Neuroscience 23 11 683--704 . APACrefD...

    2022 doi

  34. [42]

    APACrefauthors \ 1994

    mackay_bayesian_1994 APACrefauthors MacKay, D J. APACrefauthors \ 1994 . Bayesian Non - Linear Modeling for the Energy Prediction Competition Bayesian Non - Linear Modeling for the Energy Prediction Competition . ASHRAE transactions 100 2 1053--1062

    1994

  35. [43]

    APACrefauthors \ 1995

    neal_bayesian_1995 APACrefauthors Neal, R M. APACrefauthors \ 1995 . Bayesian Learning for Neural Networks Bayesian Learning for Neural Networks . Doctoral thesis, University of Toronto

    1995

  36. [44]

    APACrefauthors \ 2007

    paciorek_bayesian_2007 APACrefauthors Paciorek, C J. APACrefauthors \ 2007 . Bayesian Smoothing with Gaussian Processes Using Fourier Basis Functions in the spectralGP Package Bayesian Smoothing with Gaussian Processes Using Fourier Basis Functions in the spectralGP Package . ...

    2007 doi

  37. [45]

    \ Tobar, F

    parra_spectral_2017 APACrefauthors Parra, G. \ Tobar, F. APACrefauthors \ 2017 . Spectral Mixture Kernels for Multi - Output Gaussian Processes Spectral Mixture Kernels for Multi - Output Gaussian Processes . Advances in Neural Information Processing Systems 30 6681--6690

    2017

  38. [46]

    APACrefauthors \ 1981

    priestley_spectral_1981 APACrefauthors Priestley, M B. APACrefauthors \ 1981 . Spectral analysis and time series Spectral analysis and time series \ ( 1). Elsevier

    1981

  39. [47]

    \ Rasmussen, C E

    quinonero-candela_unifying_2005 APACrefauthors Quiñonero-Candela, J. \ Rasmussen, C E. APACrefauthors \ 2005 . A Unifying View of Sparse Approximate Gaussian Process Regression A Unifying View of Sparse Approximate Gaussian Process Regression . Journal of Machine Learning Rese...

    2005

  40. [48]

    \ Yang, J

    rao_reconciling_2021 APACrefauthors Rao, S S. \ Yang, J. APACrefauthors \ 2021 . Reconciling the Gaussian and Whittle likelihood with an application to estimation in the frequency domain Reconciling the Gaussian and Whittle likelihood with an application to estimation in the f...

    2021 doi

  41. [49]

    \ Williams, C K I

    rasmussen_gaussian_2006 APACrefauthors Rasmussen, C E. \ Williams, C K I. APACrefauthors \ 2006 . Gaussian processes for machine learning Gaussian processes for machine learning . MIT Press

    2006

  42. [50]

    , Tung, F

    rauch_maximum_1965 APACrefauthors Rauch, H E. , Tung, F. \ Striebel, C T. APACrefauthors \ 1965 . Maximum likelihood estimates of linear dynamic systems Maximum likelihood estimates of linear dynamic systems . AIAA Journal 3 8 1445--1450 . APACrefDOI doi:10.2514/3.3166 APACrefDOI

    1965 doi

  43. [51]

    \ Kailath, T

    sayed_survey_2001 APACrefauthors Sayed, A H. \ Kailath, T. APACrefauthors \ 2001 . A survey of spectral factorization methods A survey of spectral factorization methods . Numerical Linear Algebra with Applications 8 6-7 467--496 . APACrefDOI doi:10.1002/nla.250 APACrefDOI

    2001 doi

  44. [52]

    \ Scharf, L L

    schreier_statistical_2010 APACrefauthors Schreier, P J. \ Scharf, L L. APACrefauthors \ 2010 . Statistical signal processing of complex-valued data: The theory of improper and noncircular signals Statistical signal processing of complex-valued data: The theory of improper and ...

    2010

  45. [53]

    , Gokcen, E

    semedo_statistical_2020 APACrefauthors Semedo, J D. , Gokcen, E. , Machens, C K. , Kohn, A. \ Yu, B M. APACrefauthors \ 2020 . Statistical methods for dissecting interactions between brain areas Statistical methods for dissecting interactions between brain areas . Current Opin...

    2020 doi

  46. [54]

    , Zandvakili, A

    semedo_extracting_2014 APACrefauthors Semedo, J D. , Zandvakili, A. , Kohn, A. , Machens, C K. \ Yu, B M. APACrefauthors \ 2014 . Extracting Latent Structure From Multiple Interacting Neural Populations Extracting Latent Structure From Multiple Interacting Neural Populations ....

    2014

  47. [55]

    APACrefauthors \ 1982

    silverman_kernel_1982 APACrefauthors Silverman, B W. APACrefauthors \ 1982 . Kernel Density Estimation Using the Fast Fourier Transform Kernel Density Estimation Using the Fast Fourier Transform . Journal of the Royal Statistical Society Series C: Applied Statistics 31 1 93--9...

    1982 doi

  48. [56]

    \ Kohn, A

    smith_spatial_2008 APACrefauthors Smith, M A. \ Kohn, A. APACrefauthors \ 2008 . Spatial and Temporal Scales of Neuronal Correlation in Primary Visual Cortex Spatial and Temporal Scales of Neuronal Correlation in Primary Visual Cortex . Journal of Neuroscience 28 48 12591--126...

    2008 doi

  49. [57]

    \ Ghahramani, Z

    snelson_sparse_2005 APACrefauthors Snelson, E. \ Ghahramani, Z. APACrefauthors \ 2005 . Sparse Gaussian Processes using Pseudo -inputs Sparse Gaussian Processes using Pseudo -inputs . Advances in Neural Information Processing Systems 18 1257--1264

    2005

  50. [58]

    , Hensman, J

    solin_infinite-horizon_2018 APACrefauthors Solin, A. , Hensman, J. \ Turner, R E. APACrefauthors \ 2018 . Infinite- Horizon Gaussian Processes Infinite- Horizon Gaussian Processes . Advances in Neural Information Processing Systems 31 3486--3495

    2018

  51. [59]

    \ Särkkä, S

    solin_explicit_2014 APACrefauthors Solin, A. \ Särkkä, S. APACrefauthors \ 2014 . Explicit Link Between Periodic Covariance Functions and State Space Models Explicit Link Between Periodic Covariance Functions and State Space Models . Proceedings of the Seventeenth Internationa...

    2014

  52. [60]

    , Aydin, C

    steinmetz_neuropixels_2021 APACrefauthors Steinmetz, N A. , Aydin, C. , Lebedeva, A. , Okun, M. , Pachitariu, M. , Bauza, M. Harris, T D. APACrefauthors \ 2021 . Neuropixels 2.0: A miniaturized high-density probe for stable, long-term brain recordings Neuropixels 2.0: A miniat...

    2021 doi

  53. [61]

    , Zatka-Haas, P

    steinmetz_distributed_2019 APACrefauthors Steinmetz, N A. , Zatka-Haas, P. , Carandini, M. \ Harris, K D. APACrefauthors \ 2019 . Distributed coding of choice, action and engagement across the mouse brain Distributed coding of choice, action and engagement across the mouse bra...

    2019 doi

  54. [62]

    , Olhede, S C

    sykulski_debiased_2019 APACrefauthors Sykulski, A M. , Olhede, S C. , Guillaumin, A P. , Lilly, J M. \ Early, J J. APACrefauthors \ 2019 . The debiased Whittle likelihood The debiased Whittle likelihood . Biometrika 106 2 251--266 . APACrefDOI doi:10.1093/biomet/asy071 APACrefDOI

    2019 doi

  55. [63]

    APACrefauthors \ 2009

    titsias_variational_2009 APACrefauthors Titsias, M. APACrefauthors \ 2009 . Variational Learning of Inducing Variables in Sparse Gaussian Processes Variational Learning of Inducing Variables in Sparse Gaussian Processes . Proceedings of the Twelth International Conference on A...

    2009

  56. [64]

    APACrefauthors \ 2010

    turner_statistical_2010 APACrefauthors Turner, R E. APACrefauthors \ 2010 . Statistical models for natural sounds Statistical models for natural sounds . Doctoral thesis, University College London

    2010

  57. [65]

    , Carlson, D E

    ulrich_gp_2015 APACrefauthors Ulrich, K R. , Carlson, D E. , Dzirasa, K. \ Carin, L. APACrefauthors \ 2015 . GP Kernels for Cross - Spectrum Analysis GP Kernels for Cross - Spectrum Analysis . Advances in Neural Information Processing Systems 28 1999--2007

    2015

  58. [66]

    APACrefauthors \ 1951

    whittle_hypothesis_1951 APACrefauthors Whittle, P. APACrefauthors \ 1951 . Hypothesis Testing in Time Series Analysis Hypothesis Testing in Time Series Analysis . Almqvist & Wiksells boktr

    1951

  59. [67]

    , Cowley, B R

    williamson_scaling_2016 APACrefauthors Williamson, R C. , Cowley, B R. , Litwin-Kumar, A. , Doiron, B. , Kohn, A. , Smith, M A. \ Yu, B M. APACrefauthors \ 2016 . Scaling Properties of Dimensionality Reduction for Neural Populations and Network Models Scaling Properties of Dim...

    2016 doi

  60. [68]

    \ Adams, R

    wilson_gaussian_2013 APACrefauthors Wilson, A. \ Adams, R. APACrefauthors \ 2013 . Gaussian Process Kernels for Pattern Discovery and Extrapolation Gaussian Process Kernels for Pattern Discovery and Extrapolation . Proceedings of the 30th International Conference on Machine Le...

    2013

  61. [69]

    APACrefauthors \ 1927

    wirtinger_zur_1927 APACrefauthors Wirtinger, W. APACrefauthors \ 1927 . Zur formalen Theorie der Funktionen von mehr komplexen Veränderlichen Zur formalen Theorie der Funktionen von mehr komplexen Veränderlichen . Mathematische Annalen 97 357--376

    1927

  62. [70]

    , Roy, N A

    wu_gaussian_2017 APACrefauthors Wu, A. , Roy, N A. , Keeley, S. \ Pillow, J W. APACrefauthors \ 2017 . Gaussian process based nonlinear latent structure discovery in multivariate spike train data Gaussian process based nonlinear latent structure discovery in multivariate spike...

    2017

  63. [71]

    \ Yuste, R

    yang_vivo_2017 APACrefauthors Yang, W. \ Yuste, R. APACrefauthors \ 2017 . In vivo imaging of neural activity In vivo imaging of neural activity . Nature Methods 14 4 349--359 . APACrefDOI doi:10.1038/nmeth.4230 APACrefDOI

    2017 doi

  64. [72]

    , Cunningham, J P

    yu_gaussian-process_2009 APACrefauthors Yu, B M. , Cunningham, J P. , Santhanam, G. , Ryu, S I. , Shenoy, K V. \ Sahani, M. APACrefauthors \ 2009 . Gaussian- Process Factor Analysis for Low - Dimensional Single - Trial Analysis of Neural Population Activity Gaussian- Process F...

    2009

  65. [73]

    \ Kohn, A

    zandvakili_coordinated_2015 APACrefauthors Zandvakili, A. \ Kohn, A. APACrefauthors \ 2015 . Coordinated Neuronal Activity Enhances Corticocortical Communication Coordinated Neuronal Activity Enhances Corticocortical Communication . Neuron 87 4 827--839 . APACrefDOI doi:10.101...

    2015 doi

  66. [74]

    , Leithead, W

    zhang_time-series_2005 APACrefauthors Zhang, Y. , Leithead, W. \ Leith, D. APACrefauthors \ 2005 . Time-series Gaussian Process Regression Based on Toeplitz Computation of O ( N2 ) Operations and O ( N )-level Storage Time-series Gaussian Process Regression Based on Toeplitz C...

    2005

  67. [75]

    \ Park, I M

    zhao_variational_2017 APACrefauthors Zhao, Y. \ Park, I M. APACrefauthors \ 2017 . Variational Latent Gaussian Process for Recovering Single - Trial Dynamics from Population Spike Trains Variational Latent Gaussian Process for Recovering Single - Trial Dynamics from Population...

    2017 doi

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.