- Rating
- 2
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper present GenBrain. It investigating how to probabilistically predict and generate EEGs with strong randomness. It predicts the distribution parameters for each channel over the next second based on historical EEG data, constructing a parametric generative model using Rayleigh amplitudes, von Mises phases, 1/f background components, and phase-amplitude coupling. The model employs a block-causal Transformer architecture, integrating a distribution prediction head, an auxiliary waveform prediction head, and physiological prior constraints; generation involves sampling followed by an inverse FFT, with the option to incorporate residuals provided by the waveform head.Experiments involved pre-training on 14 public EEG datasets, followed by zero-shot generation evaluation on 10 datasets not included in the pre-training phase.
Critical strengths
Board evaluation on mutiple datasets and tasks: The model pre-trains on 14 EEG datasets and performs zero-shot evaluation on 10 datasets covering six scenarios. It also examines single-step prediction likelihood, distribution calibration, and autoregressive generation quality. This design tests the method's applicability across diverse EEG data conditions and mitigates the risk of conclusions relying on a single dataset or task. Such comprehensive experimental evidence makes the proposed method more convincing.
Critical weaknesses
1. The advantages over latent prediction have not been fully substantiated. Directly reconstructing highly random EEG waveforms may not be an ideal pre-training target, and this motivation is reasonable, but it can still support predictions in the learned latent space. The paper lacks a direct comparison to justify the choice of these manually designed distribution parameters, particularly against probabilistic latent prediction methods that also account for uncertainty.
2. The validation of distribution parameter dynamics and their physiological significance is insufficiently systematic. Since the manually defined distribution forms constrain the model's representational space, the ability of these parameters to effectively capture actual EEG variations is crucial to the method's validity. Although the appendix provides local analyses such as spectral slope correlation, the paper has not yet systematically shown the time variation curves of various parameters, as well as the correspondence between the predicted parameters and the actual EEG estimated parameters. Consequently, there is insufficient evidence to confirm that these parameters possess the claimed physiological interpretability.
Questions for authors
-
Latent prediction offers an alternative to directly regressing highly stochastic EEG waveforms. Could the authors further elaborate on the specific advantages of predicting parameters of a predefined physiological distribution versus predicting a learned latent representation? Would it be possible to provide a simple comparative experiment against probabilistic latent prediction.
-
Could the authors demonstrate the trajectories of predicted parameters as pre-training progresses, or report the parameter values at different training stages? Illustrating parameter changes surrounding brain state transitions would help substantiate claims regarding physiological interpretability.
-
The model is restricted to adjusting parameters within a predefined family of distributions. Have the authors considered or experimented with more flexible distribution families (i.e. Weibull distribution)?
-
In the limitations section, the authors note that the phase learning branch is weak and that there is a discrepancy between the predicted phase concentration and the observed data. Given that phase modeling is one of the core components of the proposed method, please explain why this situation was not further fixed.
9 comments
Response to Reviewer Eimr [2/N]
Weakness 2
The validation of distribution parameter dynamics and their physiological significance is insufficiently systematic. Since the manually defined distribution forms constrain the model's representational space, the ability of these parameters to effectively capture actual EEG variations is crucial to the method's validity. Although the appendix provides local analyses such as spectral slope correlation, the paper has not yet systematically shown the time variation curves of various parameters, as well as the correspondence between the predicted parameters and the actual EEG estimated parameters. Consequently, there is insufficient evidence to confirm that these parameters possess the claimed physiological interpretability.
Response: Agreed — the appendix's local analyses (e.g., the spectral-slope correlation) were a start, but a systematic validation was needed, and we built one: a trajectory analysis of the predicted physiological parameters (predicted vs. observed band amplitudes, phases, and 1/ slope over 30-s epochs) — on held-out ISRUC and CHB-MIT records we compare the predicted parameter trajectories against moment-matched references computed from the observed signal, with unsupervised brain-state segments (-means, , on per-patch band relative power; no external labels). Pooled Pearson correlations (per-record and pooled):
| Dataset | Parameter | (record 0) | (record 1) | (pooled) | |
|---|---|---|---|---|---|
| ISRUC | 0.74 | 0.53 | 0.76 | 1160 | |
| ISRUC | 0.75 | 0.68 | 0.75 | 1160 | |
| ISRUC | 0.82 | 0.71 | 0.84 | 1160 | |
| ISRUC | 0.61 | 0.64 | 0.62 | 1160 | |
| ISRUC | 0.86 | 0.58 | 0.87 | 1160 | |
| ISRUC | (1/ slope) | 0.71 | 0.53 | 0.68 | 1160 |
| ISRUC | 0.12 | 0.05 | 0.14 | 1160 | |
| CHB-MIT | 0.85 | 0.73 | 0.76 | 18 | |
| CHB-MIT | 0.34 | 0.78 | 0.56 | 18 | |
| CHB-MIT | 0.39 | 0.79 | 0.70 | 18 | |
| CHB-MIT | −0.14 | 0.63 | 0.52 | 18 | |
| CHB-MIT | −0.33 | 0.96 | 0.55 | 18 | |
| CHB-MIT | (1/ slope) | 0.10 | 0.04 | 0.18 | 18 |
| CHB-MIT | −0.13 | 0.23 | 0.07 | 18 |
Band-amplitude and slope trajectories track the observed signal (ISRUC : ); the CHB-MIT / rows are limited by the 10-s test windows (18 scored patches), the / rows show the same record-level instability there (record-0 /, pooled /), and the rows use a moment estimator with a known downward concentration bias (quantified in our response to Question 4 below). Figure 3 (App. M, p. 29) shows the predicted trajectories following sleep-state transitions of the observed signal. The figure and this table — which now covers all five band amplitudes — are included in the revision, highlighted in blue (App. M: Figure 3, p. 29; Table 14, p. 29; the CHB-MIT panel as Figure 5, p. 31).
Response to Reviewer Eimr [3/N]
Question 1
Latent prediction offers an alternative to directly regressing highly stochastic EEG waveforms. Could the authors further elaborate on the specific advantages of predicting parameters of a predefined physiological distribution versus predicting a learned latent representation? Would it be possible to provide a simple comparative experiment against probabilistic latent prediction.
Response: Yes — the conditional-VAE baseline described in our response to Weakness 1 above is exactly this comparative experiment, and it is now complete. The evaluation protocol is identical to the paper's dual model. (a) One-step likelihood on the paper's scored split (full test set): to make the comparison direct, we score the CVAE through the same family lens the paper uses for the signal-only ablation — the paper already scores the dual head (parameters predicted directly) and signal-only (parameters induced by moment-fitting the predicted waveform) with one fixed closed-form family likelihood and Energy Score — so we moment-fit the CVAE's predictive samples into physiological parameters and apply the identical scoring; the lens columns below are therefore directly comparable with the dual column, exactly as dual vs. signal-only is in Table 1. The CVAE's own waveform-space likelihood (importance sampling, ; and ELBO) is reported as a supplementary column in its native units. (b) Rollout on the full test split under the paper's per-dataset context/rollout lengths: sampled from the prior, the decoded waveform fed back into the history with block-causal re-encoding at every step, plus an RMS-rescaled variant replicating the paper's rescaling row. (c) Family-lens calibration: 200 prior samples per position moment-fit into , on which we run the paper's PIT machinery (per-band KS), compared against the dual model's published calibration numbers.
One-step predictive likelihood (full test set, paper's scored split; NLL / Energy Score):
| Dataset | Dual (family) NLL / ES | CVAE (pooled) NLL / ES | CVAE (point) NLL | Signal-only NLL / ES | CVAE waveform NLL (IS / ELBO) |
|---|---|---|---|---|---|
| ISRUC | 4.308 / 1.209 | 6.669 / 1.206 | 155.337 | 11.324 / 1.286 | −345.1 / −389.9 |
| CHB-MIT | 5.523 / 4.507 | 6.853 / 4.311 | 209.173 | 10.893 / 4.771 | −111.9 / −159.0 |
| SHU-MI | 2.932 / 0.235 | 3.105 / 0.244 | 31.709 | 4.928 / 0.478 | −568.8 / −598.3 |
| FACED | 4.777 / 0.797 | 5.463 / 0.785 | 87.385 | 8.040 / 0.877 | −342.6 / −381.7 |
| SEED-V | 4.057 / 1.277 | 4.340 / 1.253 | 89.753 | 5.759 / 1.340 | −277.9 / −317.0 |
| SEED-VII | 4.365 / 1.795 | 4.568 / 1.745 | 116.307 | 5.938 / 1.893 | −233.3 / −270.5 |
| PhysioNet-MI | 5.050 / 2.878 | 5.742 / 2.846 | 177.839 | 7.727 / 3.070 | −107.3 / −146.5 |
| BCIC-IV-2a | 4.400 / 0.736 | 5.475 / 0.845 | 58.951 | 7.393 / 0.800 | −334.7 / −380.4 |
| MODMA | 3.971 / 0.866 | 4.403 / 0.844 | 57.324 | 6.108 / 0.932 | −359.7 / −392.3 |
| MentalArithmetic | 4.511 / 0.767 | 6.680 / 0.817 | 75.125 | 9.473 / 0.834 | −331.3 / −376.1 |
| Mean | 4.390 / 1.507 | 5.330 / 1.490 | 105.890 | 7.758 / 1.628 | −301.3 / −341.2 |
(Bold = lower family NLL per row. CVAE-point = the prior-mean latent decoded to a point waveform and moment-fit, the verbatim analogue of the signal-only treatment; CVAE-pooled = 200 prior samples moment-fit per position. The last column is the CVAE's NLL under its own Gaussian waveform likelihood — strictly proper but in different units, shown only as supplementary and never differenced.)
The dual head attains the lower family NLL on 10/10 datasets (mean 4.390 vs 5.330), and the CVAE in turn beats the signal-only ablation on 10/10 (mean 5.330 vs 7.758): a learned continuous bottleneck does help over raw waveform prediction, but the physiological parameterization helps more. The sample-based Energy Score is close (CVAE 1.490 vs dual 1.507, CVAE better on 7/10), and the point-decoded variant collapses (NLL 31.709–209.173), confirming the CVAE's value is distributional rather than point-predictive.
Response to Reviewer Eimr [1/N]
Weakness 1
Response: We agree a direct comparison is the right way to settle this, and we ran the strongest form of this baseline: a conditional-VAE baseline sharing the same trunk architecture, data pipeline, and the full 10-epoch training budget as our distribution head (64-d Gaussian latent, Gaussian waveform decoder, ELBO-trained; 6.15M vs 5.81M parameters), whose reconstruction target is the raw waveform patch — it never routes through our 19-dimensional physiological parameterization — trained jointly from scratch. Design: trunk states ht are block-causal (ht encodes x≤t); a posterior head q(zt∣h≤t) over a 64-dim latent (amortized: it sees the current target, as standard), a prior head p(zt∣h<t) (used for sampling and rollout), and a decoder mapping [zt;ht−1] to a Gaussian likelihood over the 200 raw waveform dimensions (μ, σ per dim). Training is teacher-forced ELBO (Gaussian waveform NLL + KL) on the identical data pipeline, batch size, learning rate and 10-epoch budget as the paper's dual model. Capacity: trunk 5.81M (identical architecture), variational heads +0.34M, total 6.15M parameters.
Our priors, stated before the numbers: (a) the physiological parameterization gives every prediction error a mechanistic reading (band power, phase locking, aperiodic slope), whereas latent errors are unattributable; (b) it yields a closed-form likelihood and exact sampling, enabling the zero-shot cross-dataset protocol; (c) our signal-only ablation — a soft version of "learn the representation implicitly" — trails the dual model by 1.6–7.0 nats (Table 1). The complete results are given in our response to Question 1 below.