Pith. sign in

REVIEW 4 major objections 6 minor 300 references

Group EEG synchrony tracks how fast emotional arousal changes, not how strong it is, and can stand in for continuous labels.

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 · grok-4.5

2026-07-31 14:45 UTC pith:4TLCXHPN

load-bearing objection Solid multi-dataset empirical core: DNS tracks arousal change rate better than level, with real parameter guidance—useful, not revolutionary, and the annotation-free pitch runs ahead of the evidence. the 4 major comments →

arxiv 2607.28204 v1 pith:4TLCXHPN submitted 2026-07-30 cs.HC

Toward Annotation-Efficient Continuous Emotion Arousal Quantification via Group-Level EEG Dynamic Neural Synchrony

Guandong Pan , Yaqian Yang , Shi Chen , Yi Zheng , Yi Zhen , Hongwei Zheng , Shaoting Tang This is my paper
classification cs.HC
keywords EEGdynamic neural synchronyinter-subject correlationCorrCAcontinuous arousalannotation efficiencyaffective computingvalence
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.

Continuous emotion labels are expensive: raters must watch every stimulus moment by moment, and the labels do not scale. This paper argues that when many people watch the same material, the shared ups and downs of their brain activity—group-level EEG dynamic neural synchrony—already carry continuous arousal information. Across four EEG datasets with 142 people and more than 207 hours of recording, that synchrony differs by emotional valence and, more importantly, lines up better with the rate of change in arousal ratings than with the raw intensity scores. The authors map which analysis choices make the link reliable: medium time windows, a slight lead of the neural signal over the ratings, and a simple first-difference EEG feature from the main shared component. If the coupling holds after a one-time check against a modest set of ratings, later groups need only watch while EEG is recorded; the synchrony itself becomes the continuous arousal signal.

Core claim

Group-level EEG dynamic neural synchrony is an empirically validated marker of continuous emotional arousal dynamics. It carries valence-dependent structure, correlates more strongly with the first-order derivative of arousal than with raw arousal, and that coupling is stable under moderate windows (10–30 s), positive lags, and first-order-difference features from the dominant CorrCA component, as confirmed by subject-split and block-permutation controls across four datasets.

What carries the argument

Dynamic neural synchrony (DNS): sliding-window Correlated Component Analysis (CorrCA) that finds spatial filters maximizing inter-subject EEG correlation, then yields a group-level time series of shared neural coupling used as a continuous arousal proxy.

Load-bearing premise

Averaged continuous arousal ratings from independent raters are a faithful enough reference for moment-to-moment emotional change that DNS–derivative correlations can be read as capture of arousal dynamics rather than shared response to low-level audiovisual structure or annotation lag.

What would settle it

On held-out naturalistic clips, after matching or removing luminance and motion energy, test whether DNS still correlates more with the arousal derivative than with raw arousal under the stated window/lag/feature settings; if the derivative advantage and valence effects disappear, the central claim fails.

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

If this is right

  • A one-time calibration of DNS against a modest annotated subset can amortize continuous arousal labeling across later passive-viewing cohorts.
  • Practitioners can start with 10–30 s windows, first-order-difference EEG features, and the leading CorrCA component without per-study retuning.
  • Emotion quantification can target change-rate dynamics directly from group neural activity rather than only static intensity scores.
  • Positive lag of DNS ahead of ratings motivates treating group synchrony as a candidate leading signal in continuous monitoring designs.
  • Portable EEG plus passive viewing becomes a more realistic bottleneck than hundreds of rater-hours per stimulus hour.

Where Pith is reading between the lines

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

  • If DNS mainly tracks change rate, fusion models that predict both level and derivative may need less labeled data than level-only supervised pipelines.
  • The same parameterization map could be stress-tested on interactive or social settings where narrative pace no longer dominates the timescale.
  • A natural next measurement is whether audio-driven DNS alone recovers enough derivative structure to support annotation-light evaluation of soundtrack or podcast content.
  • Individual-level decoding may still need covariates; the group signal’s value is population proxy, not personal readout.

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

4 major / 6 minor

Summary. The paper proposes group-level EEG dynamic neural synchrony (DNS), computed via sliding-window CorrCA, as a population-level marker of continuous emotional arousal that could reduce reliance on per-subject manual labels. Across four datasets (SEED, SEED-IV, SEED-VII, BA VE; 142 subjects, >207 h), the authors report three main results: (i) DNS shows valence-dependent differences (positive > negative/neutral; all p<0.003); (ii) DNS correlates more strongly with the first-order derivative of continuous arousal ratings than with raw arousal; and (iii) coupling is systematically modulated by window size (best ~10–30 s), positive lags (0–10 s), FD features, and the leading CorrCA component. Block permutation, subject-split replication, and circular-shift nulls are used to argue the associations are not chance alignment or split artifacts. The manuscript frames DNS as an empirically validated, annotation-efficient surrogate for continuous arousal dynamics under naturalistic viewing.

Significance. If the DNS–arousal-derivative coupling is genuinely affective rather than largely stimulus-envelope-driven, the work would matter for affective computing and naturalistic neuroscience: it extends EEG-ISC from engagement/preference into continuous emotion dynamics, supplies the first broad parameter map for emotion-related EEG synchrony, and sketches a one-time-calibration model that could amortize costly continuous annotation. Strengths include multi-dataset scale, transparent parameter sweeps, and conservative resampling (10k block permutations; subject-split consistency). Absolute correlations remain modest, so significance hinges on careful interpretation and on whether future work can separate emotional change from low-level audiovisual structure. The methodological guidelines (window, lag, FD + Component 0) are a concrete contribution even if the full annotation-efficiency claim is not yet demonstrated.

major comments (4)
  1. [§4.2, Fig. 3–5, Limitations] §4.2–4.3 and Limitations: The central mechanistic claim that DNS “captures the rate of emotional change rather than static intensity” rests on r_deriv > r_raw against multi-rater continuous ratings. Absolute mean r_deriv is modest (~0.05–0.20), and the paper acknowledges uncontrolled luminance/motion energy and that positive lags (Fig. 5; 0–10 steps) are compatible with annotation latency as well as neural lead. Without at least one control that partials out or matches low-level AV envelopes (or a stimulus set with emotion dynamics orthogonal to cuts/loudness), the derivative advantage can still be read as shared locking to stimulus change plus rating delay. This is load-bearing for both the mechanistic sentence in the abstract and the “validated continuous-arousal marker” framing. Please either add a confound-control analysis or substantially qualify the claim to “DNS tracks change-rela
  2. [§2.3, §5.2, Abstract] §2.3 and §5.2: The annotation-efficiency operating model (“one-time validation on a calibrated subset; thereafter DNS serves as surrogate without per-study labeling”) is presented as a near-ready deployment path, but the experiments only establish concurrent coupling on already-rated material. There is no held-out transfer test (e.g., fit a DNS→arousal-derivative mapping on a subset of clips/datasets and predict continuous dynamics on unseen clips or a new cohort without using those clips’ labels beyond evaluation). Until such a test exists, “toward annotation-efficient” should remain aspirational in title/abstract/conclusion, and claims of readiness “without per-study parameter re-tuning” should be toned down to “parameter ranges that generalize across the four datasets studied.”
  3. [§3.1, Table 1, §4.2] §3.1 and Table 1: Continuous arousal for SEED and BA VE is author-collected from independent raters; SEED-VII uses built-in 0.25 Hz ratings resampled to 1 Hz. Split-half reliability is reported (SM Fig. S1), but the main text lacks sufficient protocol detail for reproducibility and for judging temporal fidelity of the derivative (rating interface, instructions, whether raters saw stimuli once or with re-exposure, temporal smoothing before differencing, alignment of the 0.25 Hz SEED-VII series). Because the headline result is specifically about first-order derivatives, the temporal properties of the reference signal are load-bearing. Please expand methods (or a dedicated SM section referenced in §3.1) on annotation acquisition, preprocessing, and how derivatives were computed and synchronized to DNS windows.
  4. [§4.3, Table 2, Abstract, §5] §4.3 Robustness / Table 2: Block permutation supports r_deriv on SEED and SEED-VII at W=30 s, but BA VE is excluded (only 3 films) and r_raw is non-significant under the same null. Subject-split “both-significant” rates are only ~50–59%. These results support non-artifactual structure but sit uneasily next to language that “establishes DNS as an empirically validated group-level marker.” Please report effect-size distributions and null comparisons more prominently in the main text (not only SM), clarify success rates, and align abstract/conclusion wording with the strength of evidence—validated coupling under specified parameters, not a drop-in continuous arousal quantifier.
minor comments (6)
  1. [§3.3] CorrCA is described clearly in §3.3, but the generalized eigenproblem is written as (R11+R22)w = λ R12 w; standard CorrCA forms often place cross-covariance on the other side of the pencil. Please confirm the equation matches the implementation and cite the exact Dmochowski convention used.
  2. [Figure 3] Figure 3 caption is dense; consider moving valence-stratified proportions of significant films into a small table for readability.
  3. [Table 1, throughout] “BA VE” appears with a space inconsistently (BA VE vs BAVE); standardize the dataset name.
  4. [§3.1, §4.3.2] Report exact sampling of continuous ratings after resampling (SEED-VII from 0.25 Hz) and whether derivative was simple first difference at 1 Hz or smoothed—critical for interpreting lag heatmaps.
  5. [§2.2] Related work could briefly contrast DNS with hyperscanning inter-brain synchrony to avoid conflation with dual-person setups (Ding et al. is cited but the distinction could be sharper).
  6. [§3.4, References] Typos/style: “ANOV A” spaced oddly in places; “o” truncation in reader notes aside, check PDF for “P ´erez” and similar accent artifacts in references.

Circularity Check

0 steps flagged

No circularity: DNS is computed from EEG alone; arousal coupling is an external empirical test, not true by construction.

full rationale

The paper’s load-bearing chain is empirical association, not a closed derivation. DNS is obtained via CorrCA on multi-subject EEG (spatial filters maximizing inter-subject covariance; sliding-window application), with no dependence on arousal labels in the definition or computation (§3.3). Continuous arousal time series are independent multi-rater behavioral references (SEED/BA VE self-collected; SEED-VII built-in), used only post hoc for Pearson coupling, valence ANOVAs, lag/window sweeps, block permutation, and subject-split checks (§3.1, §3.4, §4). Preferring the first-order derivative of arousal over raw intensity is a comparative correlation result, not a quantity fitted into DNS and then re-predicted. Parameter defaults (FD, Component 0, 10–30 s, positive lags) are selected after sweeps and validated with nulls; that is exploratory methodology, not self-definitional forcing. Author-overlapping citations (e.g., Pan et al. 2026) appear only in related work and do not underwrite uniqueness or the central DNS–arousal claims. No step reduces a claimed prediction to its own inputs by construction.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 1 invented entities

The work is empirical, not axiomatic. Load-bearing commitments are standard signal-processing and affective-science assumptions plus analysis choices (window, lag, feature, component) selected from sweeps. No new physical entity is postulated; DNS is an operational CorrCA-derived time series. Free parameters are methodological knobs that modulate reported coupling strength.

free parameters (5)
  • sliding window length W = 10–30 s (preferred operating range)
    Swept 2–40 s; reported ‘best’ coupling and valence sensitivity concentrate at 10–30 s. Central practical guideline depends on this choice.
  • temporal lag (DNS vs arousal) = 0–10 steps preferred
    Swept −20 to +20 steps; significant cells cluster at 0–10 s positive lag. Affects interpretation of lead/lag and reported coupling maps.
  • CorrCA component index = Component 0
    Components 0–4 evaluated; Component 0 retained as dominant shared pattern for main claims.
  • EEG feature type (FD vs DE bands vs raw) = First-order Difference (primary)
    FD chosen as primary after comparison; DE-gamma comparable. Coupling strength depends on this representation choice.
  • CorrCA covariance rank retention = top 10 dimensions
    Regularization keeps top 10 dimensions of R11+R22; fixed implementation choice affecting spatial filters.
axioms (5)
  • domain assumption Inter-subject temporal coupling under shared stimulation indexes stimulus-driven shared neural processing (ISC framework).
    Invoked throughout Introduction and Related Work via Hasson/Nastase/Dmochowski lineage as justification for DNS as emotion-relevant signal.
  • domain assumption Averaged continuous arousal ratings from multiple raters are a valid behavioral reference for group emotional dynamics at 1 Hz.
    Section 3.1 and annotation reliability checks; required to interpret DNS–arousal correlations as emotion coupling.
  • standard math CorrCA generalized eigenproblem on paired subject covariances yields components ordered by shared variance suitable for sliding-window DNS.
    Section 3.3; standard CorrCA reduction used without re-derivation.
  • domain assumption Block permutation across clips and subject-split re-estimation adequately control autocorrelation and group-composition artifacts.
    Section 3.4 and Robustness Validation; underwrites non-artifact claim for DNS–arousal association.
  • ad hoc to paper One-time DNS–arousal calibration on a rated subset transfers so later unlabeled groups yield usable arousal-dynamics surrogates.
    Discussion §5.2 annotation-efficiency operating model; asserted as deployment implication beyond the reported within-study correlations.
invented entities (1)
  • DNS (dynamic neural synchrony) as continuous arousal-dynamics marker no independent evidence
    purpose: Name and operationalize sliding-window CorrCA ISC as a group-level surrogate signal for continuous arousal change.
    Not a new biological object; a methodological construct built from existing CorrCA. Independent handle is correlation with external ratings and valence ANOVA, internal to this evaluation design.

pith-pipeline@v1.2.0-daily-grok45 · 16363 in / 3515 out tokens · 74623 ms · 2026-07-31T14:45:11.157053+00:00 · methodology

0 comments
read the original abstract

Continuous emotional arousal quantification remains bottlenecked by time-consuming and labor-intensive manual annotation. This work investigates group-level EEG dynamic neural synchrony (DNS) as a principled signal for continuous arousal quantification that bypasses per-subject manual labeling. Using Correlated Component Analysis (CorrCA) with sliding-window computation across four EEG datasets spanning 142 subjects and over 207 hours, we systematically evaluate DNS as a group-level marker for emotional arousal dynamics. Three key findings emerge. First, DNS exhibits significant emotion information from valence-dependent differences (all p<0.003), with positive emotions eliciting higher synchrony. Second, DNS correlates more strongly with the first-order derivative of arousal than with raw arousal values, revealing that neural synchrony captures the rate of emotional change rather than static intensity. Third, we provide the first systematic characterization of how DNS-arousal coupling depends on key methodological choices, finding that moderate windows (10-30 s), positive lags (0-10 steps), and First-order Difference feature of EEG from the dominant CorrCA component yield consistently strong coupling. Subject-split replication and block permutation tests confirm these associations are not statistical artifacts. Our findings establish DNS as an empirically validated group-level marker toward annotation-efficient continuous emotional arousal quantification.

Figures

Figures reproduced from arXiv: 2607.28204 by Guandong Pan, Hongwei Zheng, Shaoting Tang, Shi Chen, Yaqian Yang, Yi Zhen, Yi Zheng.

Figure 1
Figure 1. Figure 1: Valence-dependent differences in manually anno [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Valence-dependent DNS differences across datasets. DNS is computed using FD features with CorrCA Component 0 [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Cross-dataset DNS-arousal correlation distributions. DNS is computed using FD features with CorrCA Component 0 [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: DNS (blue) vs. arousal derivative (red) for top-10 SEED-VII films. DNS is computed using FD features with CorrCA [PITH_FULL_IMAGE:figures/full_fig_p010_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Window-lag optimization for DNS–arousal coupling (SEED-VII). DNS is computed using FD features with CorrCA [PITH_FULL_IMAGE:figures/full_fig_p010_5.png] 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

300 extracted references · 1 canonical work pages

  1. [1]

    (78 封私信 / 28 条消息) 因果关系是真实存在，还是我们认识世界的一种方法？ - 知乎 , howpublished =

  2. [2]

    微信公众平台 , howpublished =

    AAAI Workshop AIBED2024. 微信公众平台 , howpublished =

  3. [3]

    阿里离职员工发万字长文【原文】 , journal =

  4. [4]

    The Brains of High Functioning Autistic Individuals Do Not Synchronize with Those of Others - PubMed , howpublished =

  5. [5]

    Wikipedia , copyright =

    Computational Neuroscience , year =. Wikipedia , copyright =

  6. [6]

    大A进入分歧阶段 , journal =

  7. [7]

    对大A月线周期的分析 , journal =

  8. [8]

    EEG源连通性：旨在实现大脑网络在时间和空间上的高分辨率-腾讯云开发者社区-腾讯云 , howpublished =

  9. [9]

    Full Article: Affective Neuroscience of the Emotional BrainMind: Evolutionary Perspectives and Implications for Understanding Depression , howpublished =

  10. [10]

    更强的Llama 2开源，可直接商用：一夜之间，大模型格局变了 , journal =

  11. [11]

    Intersubject Correlation , howpublished =

  12. [12]

    知乎专栏 , howpublished =

    解耦表征学习. 知乎专栏 , howpublished =

  13. [13]

    Nat Neurosci：深圳先进院朱英杰团队揭示大脑中调控成瘾的``油门-刹车''重要环路机制 , journal =

  14. [14]

    Papers with Code - DeWave: Discrete EEG Waves Encoding for Brain Dynamics to Text Translation , shorttitle =

  15. [15]

    微信公众平台 , howpublished =

    PNAS. 微信公众平台 , howpublished =

  16. [16]

    ``人工智能新范式赋能脑电/脑成像研究''学术研讨会会议通知 , journal =

  17. [17]

    视频理解综述：动作识别、时序动作定位、视频Embedding , journal =

  18. [18]

    【推荐】解读机器学习基础概念：VC维的来龙去脉 , journal =

  19. [19]

    万字长文解读Scaling Law的一切，洞见LLM的未来 , journal =

  20. [20]

    我们正站在时代的十字路口（上） , journal =

  21. [21]

    我们正站在时代的十字路口（下） , journal =

  22. [22]

    因果表征学习最新综述：连接因果科学和机器学习的桥梁 , journal =

  23. [23]

    因果科学：连接统计学、机器学习与自动推理的新兴交叉领域 , journal =

  24. [24]

    因果推断：因果表征学习的CV落地 , journal =

  25. [25]

    因果推断的8个落地与前沿方向 , journal =

  26. [26]

    张钹院士：人工智能通用理论即将形成，AI革命发展势不可挡 , howpublished =

  27. [27]

    周志华中文简历 , howpublished =

  28. [28]

    祝贺丨中国人工智能学会2023年度会士增选名单揭晓 , journal =

  29. [29]

    Understanding Affective Content of Music Videos through Learned Representations , booktitle =

    Acar, Esra and Hopfgartner, Frank and Albayrak, Sahin , editor =. Understanding Affective Content of Music Videos through Learned Representations , booktitle =. 2014 , series =

    2014

  30. [30]

    2002 , month = apr, journal =

    Neural Systems for Recognizing Emotion , author =. 2002 , month = apr, journal =

    2002

  31. [31]

    全文+视频实录:Ilya Sutskever 在NeurIPS 2024的预训练终结演讲（建议收藏） , shorttitle =

    AI工作坊 , journal =. 全文+视频实录:Ilya Sutskever 在NeurIPS 2024的预训练终结演讲（建议收藏） , shorttitle =

  32. [32]

    2025 , month = may, journal =

    The Cortical Architecture Representing the Linguistic Hierarchy of the Conversational Speech , author =. 2025 , month = may, journal =

    2025

  33. [33]

    and Fonseca, Manuel J

    Alarcao, Soraia M. and Fonseca, Manuel J. , year =. Emotions Recognition Using EEG Signals: A Survey , shorttitle =. IEEE Transactions on Affective Computing , volume =

  34. [34]

    High-Level Concepts for Affective Understanding of Images , booktitle =

    Ali, Afsheen Rafaqat and Shahid, Usman and Ali, Mohsen and Ho, Jeffrey , year =. High-Level Concepts for Affective Understanding of Images , booktitle =

  35. [35]

    1984 , month = dec, journal =

    Amygdalo-Cortical Projections in the Monkey (Macaca Fascicularis) , author =. 1984 , month = dec, journal =

    1984

  36. [36]

    2011 , month = jan, journal =

    Flow of Affective Information between Communicating Brains , author =. 2011 , month = jan, journal =

    2011

  37. [37]

    2025 , journal =

    Exploring Aggregated Wav2vec 2.0 Features and Dual-Stream TDNN for Efficient Spoken Dialect Identification , author =. 2025 , journal =

    2025

  38. [38]

    2023 , month = dec, journal =

    Scaling Laws for Language Encoding Models in fMRI , author =. 2023 , month = dec, journal =

    2023

  39. [39]

    2019 , month = apr, journal =

    Speech Synthesis from Neural Decoding of Spoken Sentences , author =. 2019 , month = apr, journal =

    2019

  40. [40]

    , year =

    Ariely, Dan and Berns, Gregory S. , year =. Neuromarketing: The Hope and Hype of Neuroimaging in Business , shorttitle =. Nature Reviews Neuroscience , volume =

  41. [41]

    2022 , month = dec, journal =

    An Overview of Emotion in Artificial Intelligence , author =. 2022 , month = dec, journal =

    2022

  42. [42]

    2023 , month = jun, journal =

    Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum , author =. 2023 , month = jun, journal =

    2023

  43. [43]

    1993 , month = jul, publisher =

    A Cognitive Theory of Consciousness , author =. 1993 , month = jul, publisher =

    1993

  44. [44]

    2018 , journal =

    Gaze-Sensitive Virtual Reality Based Social Communication Platform for Individuals with Autism , author =. 2018 , journal =

    2018

  45. [45]

    2024 , month = aug, journal =

    Sisterhood Predicts Similar Neural Processing of a Film , author =. 2024 , month = aug, journal =

    2024

  46. [46]

    Data2vec: A General Framework for Self-Supervised Learning in Speech, Vision and Language , shorttitle =

    Baevski, Alexei and Hsu, Wei-Ning and Xu, Qiantong and Babu, Arun and Gu, Jiatao and Auli, Michael , year =. Data2vec: A General Framework for Self-Supervised Learning in Speech, Vision and Language , shorttitle =. 2202.03555 , primaryclass =

    Pith/arXiv arXiv

  47. [47]

    Advances in Neural Information Processing Systems , author =

    Wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations , shorttitle =. Advances in Neural Information Processing Systems , author =. 2020 , volume =

    2020

  48. [48]

    You Are an Expert Annotator

    Bagdon, Christopher and Karmalker, Prathamesh and Gurulingappa, Harsha and Klinger, Roman , year =. "You Are an Expert Annotator": Automatic Best-Worst-Scaling Annotations for Emotion Intensity Modeling , shorttitle =. 2403.17612 , primaryclass =

    Pith/arXiv arXiv

  49. [49]

    Bai, Yifan and Bao, Yiping and Chen, Guanduo and Chen, Jiahao and Chen, Ningxin and Chen, Ruijue and Chen, Yanru and Chen, Yuankun and Chen, Yutian and Chen, Zhuofu and Cui, Jialei and Ding, Hao and Dong, Mengnan and Du, Angang and Du, Chenzhuang and Du, Dikang and Du, Yulun and Fan, Yu and Feng, Yichen and Fu, Kelin and Gao, Bofei and Gao, Hongcheng and ...

    Pith/arXiv arXiv

  50. [50]

    2024 , month = oct, journal =

    Confidence Control for Efficient Behaviour in Dynamic Environments , author =. 2024 , month = oct, journal =

    2024

  51. [51]

    A Ticket for Your Thoughts: Method for Predicting Content Recall and Sales Using Neural Similarity of Moviegoers , shorttitle =

    Barnett, Samuel B and Cerf, Moran , editor =. A Ticket for Your Thoughts: Method for Predicting Content Recall and Sales Using Neural Similarity of Moviegoers , shorttitle =. 2017 , month = jun, journal =

    2017

  52. [52]

    2015 , month = jul, journal =

    Interoceptive Predictions in the Brain , author =. 2015 , month = jul, journal =

    2015

  53. [53]

    The Theory of Constructed Emotion: An Active Inference Account of Interoception and Categorization , shorttitle =

    Barrett, Lisa Feldman , year =. The Theory of Constructed Emotion: An Active Inference Account of Interoception and Categorization , shorttitle =. Social Cognitive and Affective Neuroscience , volume =

  54. [54]

    2018 , month = may, number =

    The OMG-Emotion Behavior Dataset , author =. 2018 , month = may, number =. 1803.05434 , primaryclass =

    Pith/arXiv arXiv 2018

  55. [55]

    Affective Video Content Analysis: A Multidisciplinary Insight , shorttitle =

    Baveye, Yoann and Chamaret, Christel and Dellandrea, Emmanuel and Chen, Liming , year =. Affective Video Content Analysis: A Multidisciplinary Insight , shorttitle =. IEEE Transactions on Affective Computing , volume =

  56. [56]

    LIRIS-ACCEDE: A Video Database for Affective Content Analysis , shorttitle =

    Baveye, Yoann and Dellandrea, Emmanuel and Chamaret, Christel and. LIRIS-ACCEDE: A Video Database for Affective Content Analysis , shorttitle =. 2015 , month = jan, journal =

    2015

  57. [57]

    2023年8月15日 , journal =

    Music Can Be Reconstructed from Human Auditory Cortex Activity Using Nonlinear Decoding Models , author =. 2023年8月15日 , journal =

  58. [58]

    Representation Learning: A Review and New Perspectives , shorttitle =

    Bengio, Yoshua and Courville, Aaron and Vincent, Pascal , year =. Representation Learning: A Review and New Perspectives , shorttitle =. IEEE Transactions on Pattern Analysis and Machine Intelligence , volume =

  59. [59]

    2022 , journal =

    On the Influence of Shot Scale on Film Mood and Narrative Engagement in Film Viewers , author =. 2022 , journal =

    2022

  60. [60]

    2010 , journal =

    Discovering the False Discovery Rate , author =. 2010 , journal =

    2010

  61. [61]

    2015 , month = may, journal =

    Pleasure Systems in the Brain , author =. 2015 , month = may, journal =

    2015

  62. [62]

    2023 , month = jul, journal =

    Accurate Medium-Range Global Weather Forecasting with 3D Neural Networks , author =. 2023 , month = jul, journal =

    2023

  63. [63]

    Bolton, Thomas A. W. and Freitas, Lorena G. A. and Jochaut, Delphine and Giraud, Anne-Lise and Van De Ville, Dimitri , year =. Neural Responses in Autism during Movie Watching: Inter-Individual Response Variability Co-Varies with Symptomatology , shorttitle =. NeuroImage , volume =

  64. [64]

    2020 , month = sep, journal =

    The Human Brain Reveals Resting State Activity Patterns That Are Predictive of Biases in Attitudes toward Robots , author =. 2020 , month = sep, journal =

    2020

  65. [65]

    2024 , month = jul, journal =

    Exploring Retrospective Annotation in Long-Videos for Emotion Recognition , author =. 2024 , month = jul, journal =

    2024

  66. [66]

    2023 , month = oct, journal =

    Group Synchrony for Emotion Recognition Using Physiological Signals , author =. 2023 , month = oct, journal =

    2023

  67. [67]

    and Lang, Peter J

    Bradley, Margaret M. and Lang, Peter J. , year =. Measuring Emotion: The Self-Assessment Manikin and the Semantic Differential , shorttitle =. Journal of Behavior Therapy and Experimental Psychiatry , volume =

  68. [68]

    2024 , month = feb, journal =

    A Neurophysiological Basis for Aperiodic EEG and the Background Spectral Trend , author =. 2024 , month = feb, journal =

    2024

  69. [69]

    Sparks of Artificial General Intelligence: Early Experiments with GPT-4 , shorttitle =

    Bubeck, S. Sparks of Artificial General Intelligence: Early Experiments with GPT-4 , shorttitle =. 2023 , month = mar, number =. 2303.12712 , primaryclass =

    Pith/arXiv arXiv 2023

  70. [70]

    2023 , month = jul, journal =

    Large-Scale Encoding of Emotion Concepts Becomes Increasingly Similar between Individuals from Childhood to Adolescence , author =. 2023 , month = jul, journal =

    2023

  71. [71]

    2017 , journal =

    Sentiment Analysis Is a Big Suitcase , author =. 2017 , journal =

    2017

  72. [72]

    2001 , journal =

    An fMRI Study of Personality Influences on Brain Reactivity to Emotional Stimuli , author =. 2001 , journal =

    2001

  73. [73]

    Causal Narrative Comprehension: A New Perspective for Emotion Cause Extraction , shorttitle =

    Cao, Wei and Zhang, Kun and Ruan, Shulan and Tao, Hanqing and Zhao, Sirui and Wang, Hao and Liu, Qi and Chen, Enhong , year =. Causal Narrative Comprehension: A New Perspective for Emotion Cause Extraction , shorttitle =. IEEE Transactions on Affective Computing , volume =

  74. [74]

    2023 , journal =

    Learning Users Inner Thoughts and Emotion Changes for Social Media Based Suicide Risk Detection , author =. 2023 , journal =

    2023

  75. [75]

    2024 , month = may, journal =

    Shared EEG Correlates between Non-REM Parasomnia Experiences and Dreams , author =. 2024 , month = may, journal =

    2024

  76. [76]

    2022 , month = feb, journal =

    Brains and Algorithms Partially Converge in Natural Language Processing , author =. 2022 , month = feb, journal =

    2022

  77. [77]

    2022 , month = sep, journal =

    Deep Language Algorithms Predict Semantic Comprehension from Brain Activity , author =. 2022 , month = sep, journal =

    2022

  78. [78]

    2023 , month = mar, journal =

    Evidence of a Predictive Coding Hierarchy in the Human Brain Listening to Speech , author =. 2023 , month = mar, journal =

    2023

  79. [79]

    2023 , month = dec, journal =

    An Information-Theoretic Quantification of the Content of Communication between Brain Regions , author =. 2023 , month = dec, journal =

    2023

  80. [80]

    2010年12月16日 , journal =

    Dynamic Emotional and Neural Responses to Music Depend on Performance Expression and Listener Experience , author =. 2010年12月16日 , journal =

Showing first 80 references.