Pith. sign in

REVIEW 3 major objections 4 minor 110 references

A transformer-based flow-matching network can infer the number and parameters of pulses in a transient light curve orders of magnitude faster than MCMC and nested sampling.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · deepseek-v4-flash

2026-08-01 08:20 UTC pith:6QWOEQ23

load-bearing objection A genuine trans-dimensional extension of FMPE with honest limitations, but the posterior claims need SBC or coverage before this is usable. the 3 major comments →

arxiv 2607.21134 v1 pith:6QWOEQ23 submitted 2026-07-23 astro-ph.IM

Fast Inference on Astronomical Time Series with Trans-Dimensional Flow Matching Posterior Estimation

Nina van der Meulen , Tin Had\v{z}i Veljkovi\'c , Daniela Huppenkothen , Benjamin Kurt Miller , Christoph Weniger This is my paper
classification astro-ph.IM
keywords trans-dimensional inferenceflow matchingsimulation-based inferencetime seriesfast radio burstsmagnetar burstsposterior estimationtransformer architecture
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper introduces t-FMPE, a simulation-based inference method that jointly estimates the number of burst components and their parameters in astronomical time series. It claims that a transformer-based flow-matching network, trained on simulated double-sided exponential light curves, reproduces the posterior distributions obtained with Markov chain Monte Carlo and trans-dimensional nested sampling while running orders of magnitude faster. This matters because as fast radio burst catalogs grow, traditional sampling methods become computationally infeasible. The method is amortized: once trained, each burst takes seconds to sample, and selection effects such as dead time can be folded into the simulator. The paper demonstrates qualitative agreement on simulated bursts, a magnetar X-ray burst, and three fast radio bursts.

Core claim

The central claim is that t-FMPE performs amortized trans-dimensional inference on uniformly sampled univariate light curves: a classifier returns the posterior over the number of components, and a transformer encoder defines a time-dependent vector field that maps a Gaussian base distribution to the parameter posterior conditioned on the observed light curve. The authors report qualitative agreement with MCMC reference posteriors, quantified through classifier two-sample tests (54.8% for a single-peak burst, 92% for a five-peak burst, with the difference attributable to the flow-matching posterior being broader), and agreement with trans-dimensional nested sampling on a magnetar burst (both

What carries the argument

The central object is trans-dimensional flow matching posterior estimation (t-FMPE), built from two components: a multi-scale CNN classifier that estimates the posterior over the number of components, and a transformer encoder that parameterizes a conditional flow-matching vector field. The field is trained with a conditional flow-matching loss on simulated parameter–light-curve pairs, and at inference the ordinary differential equation is integrated with 200 Euler steps. Masking of excess tokens allows the transformer to handle variable effective dimensionality, and sampling the component count from the classifier propagates uncertainty about the number of components into the parameter draw

Load-bearing premise

The load-bearing assumption is that every real burst is exactly a superposition of at most twenty double-sided exponential components with parameters inside the training priors (peak times in [0.2, 0.8], amplitudes 1–300, skewness ≥ 1, rise times in a fixed range) after the data are downsampled, padded, and background-normalized; if a burst lives outside this support, the amortized posterior is biased no matter how fast inference runs.

What would settle it

Run simulation-based calibration across the training prior: if the rank statistics of true parameters under the predicted posteriors are not uniform, the posterior is miscalibrated. Alternatively, on a set of well-resolved magnetar bursts with high signal-to-noise, compare the t-FMPE component-count posteriors against trans-dimensional nested sampling; systematic offsets in the number of inferred components or in parameter coverage would falsify the claim of unbiased amortized inference.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Rapid, amortized inference makes posterior estimation feasible for large catalogs of fast radio bursts and magnetar bursts, enabling population-level studies of pulse parameter distributions.
  • Because the method is simulation-based, instrumental effects like dead time and dispersion-measure uncertainty can be included in the simulator, yielding unbiased posteriors that are difficult to obtain with analytic likelihoods.
  • The trans-dimensional framework avoids the bias of two-step peak-detection-then-fitting, propagating uncertainty about the number of components into the parameter posteriors.
  • Inference cost stays nearly constant as dimensionality grows, whereas MCMC sampling slows down dramatically with the number of components.
  • The architecture is not limited to time series and can be adapted to any trans-dimensional inference problem that involves sequential data.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The dependence on the training simulator means the method is only as reliable as the double-sided exponential burst model; the paper itself observes excess low-amplitude components on high-SNR bursts, suggesting that a mixture of pulse shapes would reduce bias.
  • Because the classifier and transformer are trained independently, the network cannot correct the component count based on parameter-level information; end-to-end training might produce sharper posteriors but could sacrifice the explicit uncertainty propagation the paper relies on.
  • The consistent broadening of t-FMPE posteriors relative to MCMC at higher component counts hints at a potential calibration issue; running simulation-based calibration across the full prior would settle whether the posterior is genuinely conservative or overdispersed.
  • The flow-matching sampler is embarrassingly parallel, so the reported 100-samples-per-second rate is likely to improve with modern hardware or fewer integration steps, making the method usable in real-time burst detection pipelines.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper introduces t-FMPE, a transformer-based flow-matching posterior estimator for trans-dimensional inference on uniformly sampled univariate light curves. The method consists of a classifier that predicts the number of burst components N and a transformer encoder that represents a flow-matching vector field conditioned on the light curve, allowing rapid generation of posterior samples over N and the per-component parameters (peak time, amplitude, rise time, skewness). The authors train four network configurations and compare against MCMC for fixed-N simulated bursts and one FRB, and against trans-dimensional nested sampling for a magnetar burst. They report qualitative agreement with reference posteriors, C2ST scores of 54.8% (one-peak simulated burst), 92% (five-peak simulated burst) and 86% (FRB20190115B), and large speedups over MCMC and nested sampling. The paper is explicitly framed as a pilot study, with limitations acknowledged in Sections 5.1, 5.3, 5.5 and 5.6.2.

Significance. If the method performs as claimed, it would be a useful contribution to simulation-based inference for astronomical time series, particularly for data sets too large for MCMC or nested sampling. The trans-dimensional extension of flow matching is technically interesting, the use of a transformer with token masking is a sensible design, and the code and pre-processed data are publicly available. The honest discussion of limitations is a strength. However, the central claim is posterior estimation, and the current evidence for the fidelity of the learned posterior is incomplete: the two higher-dimensional comparisons show C2ST scores close to non-overlapping distributions, and the paper explicitly states that no coverage or simulation-based calibration has been performed. Because a network that always returns broad, vague distributions would pass the qualitative light-curve checks yet fail as a posterior estimator, the calibration question is load-bearing for the paper's main message.

major comments (3)
  1. [§4.1.2, §4.2.2, §5.1, §5.6.2] The central claim that t-FMPE produces posteriors in agreement with MCMC/nested-sampling posteriors is not supported by the reported quantitative tests. For the five-peak simulated burst the C2ST is 92%, and for FRB20190115B it is 86% (Figs. B1 and 18), meaning a classifier can easily distinguish the FM samples from the reference samples. The authors explain this by the FM posteriors being broader, but broadness is not sufficient: a conservative posterior can still be a valid posterior only if it has correct coverage. The paper itself states in §5.1 and §5.6.2 that no coverage or simulation-based calibration has been performed and that this is needed to confirm whether the broad FM posteriors are genuinely conservative. This is precisely the load-bearing check for the word 'posterior estimation'. I request an explicit calibration analysis, e.g. simulation-based calibration or coverage te
  2. [§4.1.3, Eq. (32)] The classifier evaluation only uses the average predicted probability vector over 2000 light curves per N_true. An average can mask systematic mis-calibration for individual observations: the classifier could be overconfident for some light curves and underconfident for others, while the ensemble average still peaks at the true N. Since the full trans-dimensional posterior is the product of the classifier and the conditional flow, per-observation calibration of p_phi(N | y) is needed. I recommend reporting the distribution of the maximum-a-posteriori N across the 2000 curves, the accuracy of the MAP estimate as a function of signal-to-noise ratio, and rank-based calibration checks for the component-number posterior. The current aggregate evaluation in Fig. 12 is insufficient to validate the trans-dimensional posterior.
  3. [§5.3, §5.5] The applicability to real data rests on assumptions that are acknowledged but not quantitatively assessed. The priors restrict peak times to [0.2, 0.8], fix K=1000 bins, and require a specified background rate; §5.5 states that any out-of-bounds peaks will go undetected and that a background mismatch leads to inaccurate inference. The observational light curves are therefore padded, downsampled, and background-normalized to fit the training distribution. This is not circular, but it means that the observational posterior estimates in §§4.2.1–4.2.2 are conditional on these pre-processing choices and on the double-sided exponential model (Eq. 1). The abstract's claim of 'unbiased posteriors in the presence of observational biases such as dead time' is not demonstrated anywhere. I ask the authors to either provide explicit tests showing robustness to these pre-processing choices, or to clea
minor comments (4)
  1. [Abstract and §5.4] The abstract states 'reaching sampling rates of 100 posterior samples per second for an 80-dimensional parameter space', while §4.1.2 reports 25,000 FM samples in 40 seconds (625 samples/s) for a 20-dimensional parameter space. Please make the speed claims consistent and specify the dimensionality and hardware in each case.
  2. [§5.4] The text describes t-FMPE as 'several orders of magnitude faster' than nested sampling, but the only direct comparison is 10 seconds versus 10 minutes, a factor of 60. This is roughly one to two orders of magnitude, so 'orders of magnitude' is an overstatement unless amortized over many bursts. Please qualify.
  3. [§4.2.2, Fig. 18] The corner plot for FRB20190115B shows the FM posterior pushing against the lower skewness bound for some components. This is noted in §5.3, but it would help to explicitly state whether any posterior mass is truncated by the prior bounds and how this affects the quoted C2ST scores.
  4. [§2.2] The RFI removal procedure and the exact definition of 'downsampled by a factor shown in table 1' are described verbally. A schematic or pseudocode would improve reproducibility, especially since the matching of observational data to the K=1000 training format is a key step.

Circularity Check

0 steps flagged

No significant circularity: t-FMPE is trained on simulations from explicit priors and validated against independently re-run MCMC / nested-sampling references; no fitted parameter is relabeled as a prediction.

full rationale

The derivation chain is self-contained. t-FMPE's training objective (Eq. 26) is the standard guided flow-matching loss: parameters θ1 and N are drawn from priors p(θ)p(N), data y are generated from the simulator p(y|θ,N), and the network learns the conditional vector field (Eqs. 21-22) plus a classifier cross-entropy. Nothing in this loop fits the network to the MCMC/nested-sampling posteriors that are later used for comparison; those references are produced independently (emcee in §3.3, magnetron re-run in §4.2.1). The in-distribution classifier check in §4.1.3 draws test light curves from the same simulator, but the paper presents this as an evaluation of the learned classifier, not as an out-of-distribution prediction. The self-citations to Huppenkothen et al. (2015) supply the component model and the magnetron benchmark; since the benchmark code was re-run with matched priors and the model is an explicit input assumption, this is not a load-bearing self-citation chain. The paper's own caveats (§5.1, §5.3, §5.5, §5.6.2) acknowledge that coverage/SBC has not been performed and that out-of-prior data are not supported; these are validation and generalization limitations, not circular steps. High C2ST values (92%, 86%) indicate the FM posteriors are not trivially identical to MCMC, which is the opposite of a construction-equivalence pattern.

Axiom & Free-Parameter Ledger

7 free parameters · 5 axioms · 0 invented entities

t-FMPE introduces no new physical entities or constants. The ledger above lists the generative-model priors and network hyperparameters that the method's validity depends on. These are all chosen by hand or computed from the prior; none are fitted to observational data.

free parameters (7)
  • Pulse-shape prior bounds = t0 in [0.2,0.8], r in [3e-3,0.1], A in [1,300], s in [1,6]
    Hand-chosen in Table 3 to make simulated light curves resemble magnetar/FRB observations; the network cannot generalize outside these ranges (Section 5.3).
  • Nmax (maximum number of components) = 5, 10, or 20 depending on network
    Truncates the discrete component prior U(1,Nmax) and fixes the transformer sequence length; observational networks use Nmax=20.
  • Background rate lambda_bkg = 0, 3, or 5
    Fixed per experiment; Section 5.5 states that a mismatch between training and test background leads to inaccurate inference.
  • sigma_min = 1e-4
    Smoothing width of the conditional Gaussian path at tau=1, needed for numerical stability (Eqs. 17-18).
  • tau sampling exponent alpha = 4/3
    Skews flow-matching time sampling toward tau near 1, following Dax et al. (2023).
  • Integration steps n_steps = 200
    Number of Euler steps for posterior sampling at inference; controls accuracy versus speed (Section 3.6.3).
  • Scaling constants = A_bar=150; theta standardization from 10,000 prior samples
    Data and target scaling for training stability (Appendix A5); computed from the prior, not fitted to observations.
axioms (5)
  • standard math Flow matching learns the marginal vector field by regressing on conditional vector fields
    The training objective (Eqs. 13, 21-22) relies on the conditional flow matching result of Lipman et al. (2023).
  • standard math Bayes/prior-simulation: p(theta,y) = p(theta) p(y|theta)
    Used in Eq. 22 and the loss in Eq. 26 to generate training pairs from priors in Table 3.
  • domain assumption Burst model: light curve = constant background + sum of double-sided exponentials
    Equations 1-2; the entire inference is over these component parameters, so the model choice is load-bearing.
  • ad hoc to paper Fixed K=1000-bin time series and peak times in [0.2,0.8]
    Training data only covers this setup; observational data are preprocessed with downsampling and padding to match (Sections 2 and 5.5).
  • domain assumption Noise model: Poisson for X-ray, Gaussian with sigma=1 for radio
    Equations 4-5 and 7-8; used both in the simulator and in reference likelihoods.

pith-pipeline@v1.3.0-alltime-deepseek · 30471 in / 10734 out tokens · 103617 ms · 2026-08-01T08:20:18.712577+00:00 · methodology

0 comments
read the original abstract

The analysis of time series plays an important part in the study of (fast) transient events, including gamma-ray bursts, magnetar bursts, fast radio bursts, and solar flares. A common approach is to decompose the time series into pulses and study the pulse characteristics, such as location and amplitude, in order to constrain physical models of the source and its environment. However, estimating both the number and characteristics of these pulses presents a trans-dimensional inference problem that traditional sampling methods such as Markov Chain Monte Carlo (MCMC) and Nested Sampling struggle to solve efficiently. Simulation-based inference methods, often incorporating machine learning techniques, provide an alternative approach when traditional approaches are insufficient. Here, we introduce trans-dimensional Flow Matching Posterior Estimation (t-FMPE) implemented on a transformer architecture capable of efficient, amortized trans-dimensional inference on uniformly sampled univariate time series data. In this initial study, we apply the method to three test cases: simulated time series with known ground-truth parameters, observational data of Fast Radio Bursts and observations of X-ray bursts from magnetars. We show that t-FMPE achieves qualitative agreement with MCMC reference posteriors, successfully reproducing parameter correlations, as quantified through classifier two-sample tests. The trained network performs inference several orders of magnitude faster than MCMC and nested sampling, reaching sampling rates of 100 posterior samples per second for an 80-dimensional parameter space. The results demonstrate potential of t-FMPE for large-scale analysis of time series datasets when traditional sampling methods become infeasible, and also enable inferring unbiased posteriors in the presence of observational biases such as dead time.

Figures

Figures reproduced from arXiv: 2607.21134 by Benjamin Kurt Miller, Christoph Weniger, Daniela Huppenkothen, Nina van der Meulen, Tin Had\v{z}i Veljkovi\'c.

Figure 1
Figure 1. Figure 1: Single noise-free burst component following equation 1. The com￾ponent has amplitude 𝐴, rise time 𝑟 and a fall time given by the product of the rise time and the skewness 𝑠. The peak-time 𝑡0 is denoted by the dotted line. Figure adapted from Huppenkothen et al. (2015) [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Example burst sampled from the simulator consisting of 𝑁 = 4 components. The ground-truth noise-free flux is shown in orange. bursts (Huppenkothen et al. 2015). A single component 𝑛 is defined as 𝜆𝑛 (𝑡) = ( 𝐴𝑒 𝑡−𝑡0 𝑟 if 𝑡 ≤ 𝑡0 𝐴𝑒− 𝑡−𝑡0 𝑟 𝑠 if 𝑡 > 𝑡0 , (1) with four free parameters: amplitude 𝐴, peak time 𝑡0, rise time 𝑟, and skewness parameter 𝑠, shown in figure 1. To get a full noise-free light curve 𝜆(𝑡)… view at source ↗
Figure 3
Figure 3. Figure 3: Flow matching concept. An initial distribution 𝑝0 is continuously transformed into distribution 𝑝1 via a time-dependent vector field, where 𝑝1 may be any distribution of interest. the EnsembleSampler is run for a tunable number of steps. After this preliminary run, walkers with relatively low likelihood (lower than the highest log probability so far, rounded down to the nearest multiple of 10) are repositi… view at source ↗
Figure 4
Figure 4. Figure 4: Difference between conditional (left) and marginal (right) trajecto￾ries, induced by the conditional and marginal vector field respectively through numerical integration. 𝑝init is shown in red, 𝑝target is shown in blue. vector field that induces the global flow from 𝑝init to 𝑝target, despite this object being intractable to compute explicitly. To illustrate their difference, figure 4 shows different trajec… view at source ↗
Figure 5
Figure 5. Figure 5: Diagram of classifier architecture. Note that the classifier uses a separate CNN encoder that is independent from the one used in token generation (figure 6). network. Therefore, the network is divided into two parts: a classi￾fier, which predicts the number of components 𝑁, and a transformer encoder that represents the corresponding vector field 𝒖 𝜙 𝜏 . These modules do not share any parameters, but are t… view at source ↗
Figure 6
Figure 6. Figure 6: Diagram showing how input tokens are created. Parameters 𝜃, time series 𝑦 and time 𝜏 are encoded and concatenated into 𝑁max tokens. Each token represents a burst component that may or may not be included in self￾attention. the 𝑛-th component at time 𝜏 𝜽𝑛, 𝜏 , flow matching time 𝜏 and the time series 𝒚. As illustrated in figure 6, this results in 𝑁max tokens, such that each token represents one burst compon… view at source ↗
Figure 7
Figure 7. Figure 7: Diagram with a high-level view of the transformers architecture, further explained in section 3.6.1.4. Contributing tokens are selected with a mask that depends on the true number of components (during training) [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Diagram of the mask that is used to select relevant tokens for an example batch of 𝐵 = 8 samples with 𝑁max = 7. Column on the right indicates the number of components provided, rows on the left indicate the corresponding selected token positions in green. shown in figure 7. During self-attention, only tokens 𝑇𝑛≤𝑁 partic￾ipate. This is done by applying a mask, illustrated in figure 8, that masks out non-con… view at source ↗
Figure 9
Figure 9. Figure 9: FM posterior predicted light curves (blue) for a simulated flux light curve (grey). The dotted line indicates the ground-truth noise-less flux, with true values 𝑡0 = 0.3, 𝑟 = 3 · 10−2 , 𝑠 = 5, 𝐴 = 100. against reference posteriors obtained through MCMC sampling and evaluate the classifier. This is followed by trans-dimensional infer￾ence results of an X-ray burst from magnetar SGR J1550-5418 which we compa… view at source ↗
Figure 10
Figure 10. Figure 10: Posterior distribution comparison between MCMC (red) and FM (blue) for the simulated flux sample with 𝑁 = 1 peak shown in figure 9. The FM model was trained on data with 𝑁max = 5. For MCMC, the true number of components was given. The contours shown are the [0.5, 1, 1.5, 2]-𝜎 contours. 𝜆𝑏𝑘𝑔 = 5. The true parameter values are 𝑡0 = 0.3, log(𝑟 ) = −1.52, 𝑠 = 5, log(𝐴) = 2 [PITH_FULL_IMAGE:figures/full_fig_p… view at source ↗
Figure 11
Figure 11. Figure 11: 100 FM (left) and MCMC (right) noise-free posterior predicted light curves for the simulated flux shown in gray with 𝑁 = 5 components. The dashed line indicates the ground-truth, noise-less flux. 100%. For the distributions in figure 10, the resulting C2ST score was 54.8%. The peak time 𝑡0, skewness 𝑠 and rise time 𝑟 exhibit clear linear relationships in the corner plot. This follows from skew and rise ti… view at source ↗
Figure 13
Figure 13. Figure 13: The mean of 1000 FM and 1006 NS posterior predicted light curves (white) with the 95% posterior credible interval (red) for an observed burst from magnetar SGR J1550-5418 (trigger ID 090122173). The observed counts were down sampled Δ𝑡 = 5 · 10−4 s → Δ𝑡 = 1 · 10−3 s and, for t-FMPE, symmetrically padded with noise to resemble training data. (a) Flow Matching (b) Nested Sampling [PITH_FULL_IMAGE:figures/f… view at source ↗
Figure 12
Figure 12. Figure 12: 𝑝𝜙 (𝑁 | 𝒚) averaged over 2000 simulated light curves 𝑦 ∼ 𝑝(𝜽) 𝑝(𝒚 | 𝜽, 𝑁true ) with 𝑁true ∈ [1, 3, 5, 8]. 𝑝𝜙 (𝑁true | 𝒚) is shown in green. components, i.e. components where the distance between consecu￾tive peaks is small. When 𝑁true is large, the probability of overlapping components increases, leading to a larger amount of ambiguous light curves and hence a broader component posterior 𝒑𝜙 (𝑁 | 𝒚). 4.2 I… view at source ↗
Figure 14
Figure 14. Figure 14: FM and NS comparison between the predicted number of compo￾nents in the magnetar burst from figure 13. (A): Predicted FM posterior over the components 𝑝𝜙 (𝑁 | 𝑦). (B): The NS approximate posterior constructed from 1006 draws using the method from Huppenkothen et al. (2015), with adjusted priors to match those used in FM. (2015) was re-run for this burst with the original magnetron code, but with adjusted … view at source ↗
Figure 15
Figure 15. Figure 15: Left: The mean of 1000 posterior predicted light curves (white) with the 95% posterior credible interval (red) plotted over pre-processed radio flux of FRB20190115B (black). Right: 𝑝𝜙 (𝑁 | 𝑦), with 𝑦 the observed flux. The posterior over the components is well-constrained near 𝑁 = 3 [PITH_FULL_IMAGE:figures/full_fig_p013_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: As in figure 15, but for FRB2019124F [PITH_FULL_IMAGE:figures/full_fig_p013_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: As in figure 15, but for FRB20190122C. bution, since the network is configured to detect at most 𝑁max = 20 peaks in a light curve. Taking this into account, the two methods appear to show good agreement. 4.2.2 FRBs t-FMPE was applied to the three multi-peaked radio bursts described in section 2. Here, we use the same network configuration as in the previous section, now trained on simulated FRB profiles w… view at source ↗
Figure 18
Figure 18. Figure 18: FM (blue) and MCMC (red) posterior distributions for FRB20190115B when enforcing 𝑁 = 3. The FM posterior is broader compared to MCMC. The distributions of the peak time and rise time of the second peak (𝑡 2 0 and log 𝑟2) appear to be slightly off-set. 5.1 Broadness of t-FMPE posteriors As discussed in the previous section, we noticed that the t-FMPE posteriors tend to broaden relative to the reference pos… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

110 extracted references · 5 canonical work pages

  1. [1]

    2024 , isbn =

    Jay Alammar and Maarten Grootendorst , title =. 2024 , isbn =

    2024

  2. [2]

    Maistrello and R

    M. Maistrello and R. Maccary and C. Guidorzi , keywords =. FAST-MEPSA: An optimised and faster version of peak detection algorithm MEPSA , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.ascom.2025.101040 , url =

    arXiv 2026

  3. [3]

    2023 , eprint=

    Transformers for scientific data: a pedagogical review for astronomers , author=. 2023 , eprint=

    2023

  4. [4]

    Publications of the Astronomical Society of the Pacific , keywords =

    emcee: The MCMC Hammer. Publications of the Astronomical Society of the Pacific , keywords =. doi:10.1086/670067 , archivePrefix =. 1202.3665 , primaryClass =

    Pith/arXiv arXiv

  5. [5]

    Ensemble samplers with affine invariance , volume =

    Goodman, Jonathan and Weare, Jonathan , year =. Ensemble samplers with affine invariance , volume =. Communications in Applied Mathematics and Computational Science , doi =

  6. [6]

    and Hogg, David W

    Huppenkothen, Daniela and Brewer, Brendon J. and Hogg, David W. and Murray, Iain and Frean, Marcus and Elenbaas, Chris and Watts, Anna L. and Levin, Yuri and Horst, Alexander J. van der and Kouveliotou, Chryssa , title =. The Astrophysical Journal , abstract =. 2015 , month =. doi:10.1088/0004-637X/810/1/66 , url =

    doi:10.1088/0004-637x/810/1/66 2015

  7. [7]

    , year =

    Lorimer, D and Bailes, M and McLaughlin, M and Narkevic, D and Crawford, F. , year =. A Bright Millisecond Radio Burst of Extragalactic Origin , volume =. Science (New York, N.Y.) , doi =

  8. [8]

    and Bailes, Matthew and Ekers, Ronald and Macquart, Jean-Pierre and Crawford III, Fronefield , year=

    Burke-Spolaor, S. and Bailes, Matthew and Ekers, Ronald and Macquart, Jean-Pierre and Crawford III, Fronefield , year=. Radio Bursts with Extragalactic Spectral Characteristics Show Terrestrial Origins , volume=. The Astrophysical Journal , publisher=. doi:10.1088/0004-637x/727/1/18 , number=

    doi:10.1088/0004-637x/727/1/18

  9. [9]

    Champion, D. J. and Petroff, E. and Kramer, M. and Keith, M. J. and Bailes, M. and Barr, E. D. and Bates, S. D. and Bhat, N. D. R. and Burgay, M. and Burke-Spolaor, S. and Flynn, C. M. L. and Jameson, A. and Johnston, S. and Ng, C. and Levin, L. and Possenti, A. and Stappers, B. W. and van Straten, W. and Thornton, D. and Tiburzi, C. and Lyne, A. G. , tit...

    doi:10.1093/mnrasl/slw069 2016

  10. [10]

    Andersen, B. C. and Bandura, K. M. and Bhardwaj, M. and Bij, A. and Boyce, M. M. and Boyle, P. J. and Brar, C. and Cassanelli, T. and Chawla, P. and Chen, T. and Cliche, J.-F. and Cook, A. and Cubranic, D. and Curtin, A. P. and Denman, N. T. and Dobbs, M. and Dong, F. Q. and Fandino, M. and Fonseca, E. and Gaensler, B. M. and Giri, U. and Good, D. C. and ...

  11. [11]

    A burst storm from the repeating FRB 20200120E in an M81 globular cluster , volume=

    Nimmo, K and Hessels, J W T and Snelders, M P and Karuppusamy, R and Hewitt, D M and Kirsten, F and Marcote, B and Bach, U and Bansod, A and Barr, E D and Behrend, J and Bezrukovs, V and Buttaccio, S and Feiler, R and Gawroński, M P and Lindqvist, M and Orbidans, A and Puchalska, W and Wang, N and Winchen, T and Wolak, P and Wu, J and Yuan, J , year=. A b...

  12. [12]

    Hessels, J. W. T. and Spitler, L. G. and Seymour, A. D. and Cordes, J. M. and Michilli, D. and Lynch, R. S. and Gourdji, K. and Archibald, A. M. and Bassa, C. G. and Bower, G. C. and Chatterjee, S. and Connor, L. and Crawford, F. and Deneva, J. S. and Gajjar, V. and Kaspi, V. M. and Keimpema, A. and Law, C. J. and Marcote, B. and McLaughlin, M. A. and Par...

    2019

  13. [13]

    Universe , keywords =

    Probing the Universe with Fast Radio Bursts. Universe , keywords =. doi:10.3390/universe7040085 , adsurl =

    doi:10.3390/universe7040085

  14. [14]

    2025 , eprint=

    Morphology of 35 Repeating Fast Radio Burst Sources at Microsecond Time Scales with CHIME/FRB , author=. 2025 , eprint=

    2025

  15. [15]

    and Cook, Rick and Tomsick, John and Schmid, Christian and Grefenstette, Brian W

    Bachetti, Matteo and Harrison, Fiona A. and Cook, Rick and Tomsick, John and Schmid, Christian and Grefenstette, Brian W. and Barret, Didier and Boggs, Steven E. and Christensen, Finn E. and Craig, William W. and Fabian, Andrew C. and Fürst, Felix and Gandhi, Poshak and Hailey, Charles J. and Kara, Erin and Maccarone, Thomas J. and Miller, Jon M. and Pott...

  16. [16]

    Monthly Notices of the Royal Astronomical Society , volume =

    Beniamini, Paz and Kumar, Pawan , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2020 , month =. doi:10.1093/mnras/staa2489 , url =

    doi:10.1093/mnras/staa2489 2020

  17. [17]

    and Bandura, Kevin and Berger, Sabrina and Bhardwaj, Mohit and Boyce, Michelle M

    Amiri, Mandana and Andersen, Bridget C. and Bandura, Kevin and Berger, Sabrina and Bhardwaj, Mohit and Boyce, Michelle M. and Boyle, P. J. and Brar, Charanjot and Breitman, Daniela and Cassanelli, Tomas and Chawla, Pragya and Chen, Tianyue and Cliche, J.-F. and Cook, Amanda and Cubranic, Davor and Curtin, Alice P. and Deng, Meiling and Dobbs, Matt and (Ad...

  18. [18]

    The frontier of simulation-based inference , volume=

    Cranmer, Kyle and Brehmer, Johann and Louppe, Gilles , year=. The frontier of simulation-based inference , volume=. Proceedings of the National Academy of Sciences , publisher=. doi:10.1073/pnas.1912789117 , number=

    doi:10.1073/pnas.1912789117

  19. [19]

    2023 , eprint=

    Flow Matching for Scalable Simulation-Based Inference , author=. 2023 , eprint=

    2023

  20. [20]

    2023 , eprint=

    Flow Matching for Generative Modeling , author=. 2023 , eprint=

    2023

  21. [21]

    2024 , eprint=

    Flow Matching Guide and Code , author=. 2024 , eprint=

    2024

  22. [22]

    2010 , eprint=

    Diffusive Nested Sampling , author=. 2010 , eprint=

    2010

  23. [23]

    Journal of Statistical Software , author=

    DNest4: Diffusive Nested Sampling in C++ and Python , volume=. Journal of Statistical Software , author=. 2018 , pages=. doi:10.18637/jss.v086.i07 , abstract=

    doi:10.18637/jss.v086.i07 2018

  24. [24]

    2024 , eprint=

    Morphology of 137 Fast Radio Bursts down to Microseconds Timescales from The First CHIME/FRB Baseband Catalog , author=. 2024 , eprint=

    2024

  25. [25]

    2015 , eprint=

    MEPSA: a flexible peak search algorithm designed for uniformly spaced time series , author=. 2015 , eprint=

    2015

  26. [26]

    Magnetron: Fitting bursts from magnetars

  27. [27]

    A&A , year =

    A search for prompt -ray counterparts to fast radio bursts in the Insight-HXMT data , DOI=. A&A , year =

  28. [28]

    and Metzger, B

    Margutti, R. and Metzger, B. and Chornock, R. and Vurm, I. and Roth, Nathaniel and Grefenstette, Brian and Savchenko, Volodymyr and Cartier, R. and Steiner, J. and Terreran, Giacomo and Margalit, B. and Migliori, G. and Milisavljevic, D. and Alexander, K. and Bietenholz, M. and Blanchard, P. and Bozzo, E. and Brethauer, D. and Chilingarian, I. and Paterso...

  29. [29]

    10.1051/0004-6361/202450666

    Distribution of the number of peaks within a long gamma-ray burst: The full Fermi/GBM catalogue , DOI= "10.1051/0004-6361/202450666", url= "https://doi.org/10.1051/0004-6361/202450666", journal =

    work page doi:10.1051/0004-6361/202450666

  30. [30]

    10.1051/0004-6361/202452326

    A search for periodic activity in multi-peaked long gamma-ray bursts , DOI= "10.1051/0004-6361/202452326", url= "https://doi.org/10.1051/0004-6361/202452326", journal =

    work page doi:10.1051/0004-6361/202452326

  31. [31]

    2007 , eprint=

    Hyperflares of SGRs as an engine for millisecond extragalactic radio bursts , author=. 2007 , eprint=

    2007

  32. [32]

    Fast Radio Bursts and Their High-energy Counterparts from Magnetar Magnetospheres , volume=

    Yang, Yuan-Pei and Zhang, Bing , year=. Fast Radio Bursts and Their High-energy Counterparts from Magnetar Magnetospheres , volume=. The Astrophysical Journal , publisher=. doi:10.3847/1538-4357/ac14b5 , number=

    doi:10.3847/1538-4357/ac14b5

  33. [33]

    Pair Separation in Parallel Electric Field in Magnetar Magnetosphere and Narrow Spectra of Fast Radio Bursts , volume=

    Yang, Yuan-Pei and Zhu, Jin-Ping and Zhang, Bing and Wu, Xue-Feng , year=. Pair Separation in Parallel Electric Field in Magnetar Magnetosphere and Narrow Spectra of Fast Radio Bursts , volume=. The Astrophysical Journal Letters , publisher=. doi:10.3847/2041-8213/abb535 , number=

    doi:10.3847/2041-8213/abb535 2041

  34. [34]

    Katz, J. I. , year=. HOW SOFT GAMMA REPEATERS MIGHT MAKE FAST RADIO BURSTS , volume=. The Astrophysical Journal , publisher=. doi:10.3847/0004-637x/826/2/226 , number=

    doi:10.3847/0004-637x/826/2/226

  35. [35]

    2024 , eprint=

    Updating the first CHIME/FRB catalog of fast radio bursts with baseband data , author=. 2024 , eprint=

    2024

  36. [36]

    Monthly Notices of the Royal Astronomical Society , volume =

    Zhu-Ge, Jia-Ming and Luo, Jia-Wei and Zhang, Bing , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2022 , month =. doi:10.1093/mnras/stac3599 , url =

    doi:10.1093/mnras/stac3599 2022

  37. [37]

    2025 , eprint=

    Representation learning for fast radio burst dynamic spectra , author=. 2025 , eprint=

    2025

  38. [38]

    Júnior and Jéferson A.S

    Ailton J.B. Júnior and Jéferson A.S. Fortunato and Leonardo J. Silvestre and Thonimar V. Alencar and Wiliam S. Hipólito-Ricaldi , keywords =. Comparative analysis of machine learning techniques for feature selection and classification of Fast Radio Bursts , journal =. 2026 , issn =. doi:https://doi.org/10.1016/j.jheap.2025.100449 , url =

    arXiv 2026

  39. [39]

    2025 , eprint=

    Revised classification of the CHIME fast radio bursts with machine learning , author=. 2025 , eprint=

    2025

  40. [40]

    10.1051/0004-6361/202450953

    Comprehensive analysis of the Apertif fast radio burst sample - Similarities with young energetic neutron stars , DOI= "10.1051/0004-6361/202450953", url= "https://doi.org/10.1051/0004-6361/202450953", journal =

    doi:10.1051/0004-6361/202450953

  41. [41]

    Nested Sampling. Bayesian Inference and Maximum Entropy Methods in Science and Engineering: 24th International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering , year = 2004, editor =. doi:10.1063/1.1835238 , adsurl =

    doi:10.1063/1.1835238 2004

  42. [42]

    Traffic sign recognition with multi-scale Convolutional Networks , year=

    Sermanet, Pierre and LeCun, Yann , booktitle=. Traffic sign recognition with multi-scale Convolutional Networks , year=

  43. [43]

    Sensors , author=

    A review on multiscale-deep-learning applications , volume=. Sensors , author=. 2022 , pages=. doi:10.3390/s22197384 , number=

    work page doi:10.3390/s22197384 2022

  44. [44]

    Hastings, W. K. , title =. Biometrika , volume =. 1970 , month =. doi:10.1093/biomet/57.1.97 , url =

    doi:10.1093/biomet/57.1.97 1970

  45. [45]

    2020 , eprint=

    A Conceptual Introduction to Markov Chain Monte Carlo Methods , author=. 2020 , eprint=

    2020

  46. [46]

    New FRB discoveries and their follow-up

    The SUrvey for Pulsars and Extragalactic Radio Bursts - II. New FRB discoveries and their follow-up. Monthly Notices of the Royal Astronomical Society , keywords =. doi:10.1093/mnras/stx3074 , archivePrefix =. 1711.08110 , primaryClass =

    Pith/arXiv arXiv

  47. [47]

    and Stappers, B

    Thornton, D. and Stappers, B. and Bailes, M. and Barsdell, B. and Bates, S. and Bhat, N. D. R. and Burgay, M. and Burke-Spolaor, S. and Champion, D. J. and Coster, P. and D’Amico, N. and Jameson, A. and Johnston, S. and Keith, M. and Kramer, M. and Levin, L. and Milia, S. and Ng, C. and Possenti, A. and van Straten, W. , year=. A Population of Fast Radio ...

    doi:10.1126/science.1236789

  48. [48]

    The Astrophysical Journal , keywords =

    Fast Radio Burst Discovered in the Arecibo Pulsar ALFA Survey. The Astrophysical Journal , keywords =. doi:10.1088/0004-637X/790/2/101 , archivePrefix =. 1404.2934 , primaryClass =

    Pith/arXiv arXiv

  49. [49]

    The Astrophysical Journall , keywords =

    The Detection of an Extremely Bright Fast Radio Burst in a Phased Array Feed Survey. The Astrophysical Journall , keywords =. doi:10.3847/2041-8213/aa71ff , archivePrefix =. 1705.07581 , primaryClass =

    Pith/arXiv arXiv 2041

  50. [50]

    and Keane, E

    Petroff, E. and Keane, E. F. and Barr, E. D. and Reynolds, J. E. and Sarkissian, J. and Edwards, P. G. and Stevens, J. and Brem, C. and Jameson, A. and Burke-Spolaor, S. and Johnston, S. and Bhat, N. D. R. and Kudale, P. Chandra S. and Bhandari, S. , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2015 , month =. doi:10.1093/mnras/s...

    doi:10.1093/mnras/stv1242 2015

  51. [51]

    and Hessels, J

    Nimmo, K. and Hessels, J. and Kirsten, Franz and Keimpema, Aard and Cordes, James and Snelders, M. and Hewitt, D. and Karuppusamy, Rajagopal and Archibald, Anne and Bezrukovs, Vladislavs and Bhardwaj, Mohit and Blaauw, R. and Buttaccio, Salvo and Cassanelli, T. and Conway, J. and Corongiu, A. and Feiler, Roman and Fonseca, Esfrit and Forssén, O. and Yuan,...

  52. [52]

    and Fender, R

    Pietka, M. and Fender, R. P. and Keane, E. F. , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2014 , month =. doi:10.1093/mnras/stu2335 , url =

    doi:10.1093/mnras/stu2335 2014

  53. [53]

    The Astrophysical Journal , keywords =

    CHIME/FRB Discovery of 25 Repeating Fast Radio Burst Sources. The Astrophysical Journal , keywords =. doi:10.3847/1538-4357/acc6c1 , archivePrefix =. 2301.08762 , primaryClass =

    Pith/arXiv arXiv

  54. [54]

    and Weltman, A

    Platts, E. and Weltman, A. and Walters, A. and Tendulkar, S.P. and Gordin, J.E.B. and Kandhai, S. , year=. A living theory catalogue for fast radio bursts , volume=. doi:10.1016/j.physrep.2019.06.003 , journal=

    doi:10.1016/j.physrep.2019.06.003 2019

  55. [55]

    10.1051/0004-6361/201936284

    A cumulative search for hard X/ -ray emission associated with fast radio bursts in Fermi/GBM data , DOI= "10.1051/0004-6361/201936284", url= "https://doi.org/10.1051/0004-6361/201936284", journal =

    work page doi:10.1051/0004-6361/201936284

  56. [56]

    2023 , eprint=

    All sky archival search for FRB high energy counterparts with Swift and Fermi , author=. 2023 , eprint=

    2023

  57. [57]

    Bradley and Burns, Eric and Goldstein, Adam and Lien, Amy and Kocevski, Daniel and Briggs, Michael and Connaughton, Valerie and Miller, M

    Cunningham, Virginia and Cenko, S. Bradley and Burns, Eric and Goldstein, Adam and Lien, Amy and Kocevski, Daniel and Briggs, Michael and Connaughton, Valerie and Miller, M. Coleman and Racusin, Judith and Stanbro, Matthew , year=. A Search for High-energy Counterparts to Fast Radio Bursts , volume=. The Astrophysical Journal , publisher=. doi:10.3847/153...

    doi:10.3847/1538-4357/ab2235

  58. [58]

    Monthly Notices of the Royal Astronomical Society , volume =

    Caleb, M and Stappers, B W and Rajwade, K and Flynn, C , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2019 , month =. doi:10.1093/mnras/stz386 , url =

    doi:10.1093/mnras/stz386 2019

  59. [59]

    Monthly Notices of the Royal Astronomical Society , volume =

    Cui, Xiang-Han and Zhang, Cheng-Min and Wang, Shuang-Qiang and Zhang, Jian-Wei and Li, Di and Peng, Bo and Zhu, Wei-Wei and Wang, Na and Strom, Richard and Ye, Chang-Qing and Wang, De-Hua and Yang, Yi-Yan , title =. Monthly Notices of the Royal Astronomical Society , volume =. 2020 , month =. doi:10.1093/mnras/staa3351 , url =

    doi:10.1093/mnras/staa3351 2020

  60. [60]

    and Prochaska, J

    Macquart, J.-P. and Prochaska, J. X. and McQuinn, M. and Bannister, K. W. and Bhandari, S. and Day, C. K. and Deller, A. T. and Ekers, R. D. and James, C. W. and Marnoch, L. and Osłowski, S. and Phillips, C. and Ryder, S. D. and Scott, D. R. and Shannon, R. M. and Tejos, N. , year=. A census of baryons in the Universe from localized fast radio bursts , vo...

    doi:10.1038/s41586-020-2300-2

  61. [61]

    The Astrophysical Journal , abstract =

    Ravi, Vikram , title =. The Astrophysical Journal , abstract =. 2019 , month =. doi:10.3847/1538-4357/aafb30 , url =

    doi:10.3847/1538-4357/aafb30 2019

  62. [62]

    2023 , eprint=

    Dense Forests of Microshots in Bursts from FRB 20220912A , author=. 2023 , eprint=

    2023

  63. [64]

    Science China Physics, Mechanics, and Astronomy , keywords =

    Commissioning progress of the FAST. Science China Physics, Mechanics, and Astronomy , keywords =. doi:10.1007/s11433-018-9376-1 , archivePrefix =. 1903.06324 , primaryClass =

    Pith/arXiv arXiv 1903

  64. [65]

    The Australian square-kilometre-array pathfinder

    Science with ASKAP. The Australian square-kilometre-array pathfinder. Experimental Astronomy , keywords =. doi:10.1007/s10686-008-9124-7 , archivePrefix =. 0810.5187 , primaryClass =

    Pith/arXiv arXiv

  65. [66]

    The Astrophysical Journal Supplement , keywords =

    The Very Large Array. The Astrophysical Journal Supplement , keywords =. doi:10.1086/190688 , adsurl =

    doi:10.1086/190688

  66. [68]

    and Pleunis, Z

    Fonseca, E. and Pleunis, Z. and Breitman, D. and Sand, K. R. and Kharel, B. and Boyle, P. J. and Brar, C. and Giri, U. and Kaspi, V. M. and Masui, K. W. and Meyers, B. W. and Patel, C. and Scholz, P. and Smith, K. , title =. The Astrophysical Journal Supplement Series , abstract =. 2024 , month =. doi:10.3847/1538-4365/ad27d6 , url =

    doi:10.3847/1538-4365/ad27d6 2024

  67. [69]

    The Astrophysical Journal , keywords =

    Local Circumnuclear Magnetar Solution to Extragalactic Fast Radio Bursts. The Astrophysical Journal , keywords =. doi:10.1088/0004-637X/807/2/179 , archivePrefix =. 1501.01341 , primaryClass =

    Pith/arXiv arXiv

  68. [70]

    Annual Review of Astronomy and Astrophysics , keywords =

    Magnetars. Annual Review of Astronomy and Astrophysics , keywords =. doi:10.1146/annurev-astro-081915-023329 , archivePrefix =. 1703.00068 , primaryClass =

    Pith/arXiv arXiv

  69. [71]

    Genetics , volume =

    Beaumont, Mark A and Zhang, Wenyang and Balding, David J , title =. Genetics , volume =. 2002 , month =. doi:10.1093/genetics/162.4.2025 , url =

    doi:10.1093/genetics/162.4.2025 2002

  70. [72]

    2021 , eprint=

    Benchmarking Simulation-Based Inference , author=. 2021 , eprint=

    2021

  71. [73]

    2017 , eprint=

    Likelihood-free inference via classification , author=. 2017 , eprint=

    2017

  72. [74]

    2023 , eprint=

    Group equivariant neural posterior estimation , author=. 2023 , eprint=

    2023

  73. [75]

    2021 , eprint=

    Normalizing Flows for Probabilistic Modeling and Inference , author=. 2021 , eprint=

    2021

  74. [76]

    Tabak, E. G. and Turner, Cristina V. , title =. Communications on Pure and Applied Mathematics , volume =. doi:https://doi.org/10.1002/cpa.21423 , url =. https://onlinelibrary.wiley.com/doi/pdf/10.1002/cpa.21423 , abstract =

    doi:10.1002/cpa.21423

  75. [77]

    Proceedings of the 32nd International Conference on Machine Learning , pages =

    Variational Inference with Normalizing Flows , author =. Proceedings of the 32nd International Conference on Machine Learning , pages =. 2015 , editor =

    2015

  76. [78]

    2023 , eprint=

    Inferring Atmospheric Properties of Exoplanets with Flow Matching and Neural Importance Sampling , author=. 2023 , eprint=

    2023

  77. [79]

    Learning Summary Statistic for Approximate Bayesian Computation via Deep Neural Network , ISSN=

    Wong, Wing and Jiang, Bai and Wu, Tung-yu and Zheng, Charles , year=. Learning Summary Statistic for Approximate Bayesian Computation via Deep Neural Network , ISSN=. doi:10.5705/ss.202015.0340 , journal=

    arXiv

  78. [80]

    2025 , eprint=

    An Introduction to Flow Matching and Diffusion Models , author=. 2025 , eprint=

    2025

  79. [81]

    2020 , eprint=

    Fourier Features Let Networks Learn High Frequency Functions in Low Dimensional Domains , author=. 2020 , eprint=

    2020

  80. [82]

    2017 , eprint=

    Adam: A Method for Stochastic Optimization , author=. 2017 , eprint=

    2017

Showing first 80 references.