pith. sign in

arxiv: 2606.07707 · v1 · pith:PS6LGRWCnew · submitted 2026-06-05 · 💻 cs.LG

Decoding Naturalistic Emotion Dynamics from the Brain: An LLM-Enhanced Regression Framework

Lemei Zhang , Peng Liu , Hans Dahle Kvadsheim , August S{\ae}tre Aasv{\ae}r , Shuer Ye , Reza Bonyadi , Maryam Ziaei , Jon Atle Gulla This is my paper

Pith reviewed 2026-06-27 22:26 UTC · model grok-4.3

classification 💻 cs.LG
keywords emotion decodingdynamic functional connectivitymulti-target regressionfMRInaturalistic stimuliLLM sentiment profilesgraph XAIpsychological constructionism
0
0 comments X

The pith

Dynamic functional connectivity snapshots from fMRI track continuous, overlapping emotional trajectories more accurately than static region amplitudes.

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

The paper reframes emotion decoding as a multi-target regression task that follows several emotional dimensions as continuous trajectories over time instead of assigning single discrete labels. Large language models supply the target variables by producing fine-grained sentiment profiles from an auditory narrative, which are then aligned with human fMRI recordings. Regularized and kernel-based estimators trained on temporal snapshots of dynamic functional connectivity outperform models that use only static region-of-interest amplitude values. Graph-theoretical explainable AI methods subsequently expose emotion-specific network topologies that support distributed rather than strictly localized accounts of affect.

Core claim

Regularized and kernel-based machine learning algorithms applied to temporal snapshots of dynamic functional connectivity can continuously estimate the magnitude of multiple overlapping emotional dimensions during naturalistic narrative listening, outperforming static ROI representations and revealing emotion-specific topological configurations through graph-theoretical XAI.

What carries the argument

Temporal snapshots of Dynamic Functional Connectivity (DFC) supplied as input features to continuous regression estimators whose targets are LLM-derived sentiment profiles.

If this is right

  • Continuous multi-dimensional emotion trajectories become decodable from brain data under rapidly changing naturalistic input.
  • Distributed network dynamics explain more variance in emotional states than location-specific amplitude measures.
  • Graph XAI techniques can isolate interpretable, emotion-specific topological configurations in macroscale brain networks.
  • LLM-automated annotation supplies scalable labels for affective neuroscience experiments that use naturalistic stimuli.

Where Pith is reading between the lines

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

  • The same regression approach could be tested on other continuous cognitive or perceptual states beyond emotion.
  • Superior performance of DFC patterns might generalize to predicting behavioral or physiological markers outside the scanner.
  • Network configurations identified by the XAI step could be examined for stability across different narratives or participant groups.

Load-bearing premise

Fine-grained continuous sentiment profiles extracted by LLMs from the narrative serve as accurate proxies for the subjective emotional experiences of the human participants.

What would settle it

If models using DFC snapshots fail to outperform static ROI models when both are validated against actual continuous self-ratings collected from the same participants listening to the narrative.

Figures

Figures reproduced from arXiv: 2606.07707 by August S{\ae}tre Aasv{\ae}r, Hans Dahle Kvadsheim, Jon Atle Gulla, Lemei Zhang, Maryam Ziaei, Peng Liu, Reza Bonyadi, Shuer Ye.

Figure 1
Figure 1. Figure 1: Overview of the proposed pipeline linking dynamic functional connectivity (DFC) with emotion [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Pipeline for generating DFC datasets. After segmentation of the fMRI-data, ROIs are aligned [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Explanation for Random Forrest Regressor for Joy and Anticipation [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Topological MST projection and graph metrics of the Ridge prediction for decoded [PITH_FULL_IMAGE:figures/full_fig_p015_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Heatmaps and graph visualizations of Jaccard distances between emotional trees for the [PITH_FULL_IMAGE:figures/full_fig_p017_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Instructions for the human evaluation of emotion labelling. [PITH_FULL_IMAGE:figures/full_fig_p028_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Emotional scores for each segment in the stimuli of Alice in Wonderland chapter 1. [PITH_FULL_IMAGE:figures/full_fig_p029_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Linear Regression ROI Importance Maps for the eight emotion categories. [PITH_FULL_IMAGE:figures/full_fig_p030_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Lasso ROI Importance Maps for the eight emotion categories. [PITH_FULL_IMAGE:figures/full_fig_p031_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Ridge ROI Importance Maps for the eight emotion categories. [PITH_FULL_IMAGE:figures/full_fig_p031_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Linear SVR ROI Importance Maps for the eight emotion categories. [PITH_FULL_IMAGE:figures/full_fig_p032_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: RFR ROI Importance Maps for the eight emotion categories. [PITH_FULL_IMAGE:figures/full_fig_p032_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p033_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p034_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p035_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p036_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p037_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p038_18.png] view at source ↗
Figure 19
Figure 19. Figure 19: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p039_19.png] view at source ↗
Figure 20
Figure 20. Figure 20: Maximum spanning trees of Linear model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p040_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p041_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p042_22.png] view at source ↗
Figure 23
Figure 23. Figure 23: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p043_23.png] view at source ↗
Figure 24
Figure 24. Figure 24: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p044_24.png] view at source ↗
Figure 25
Figure 25. Figure 25: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p045_25.png] view at source ↗
Figure 26
Figure 26. Figure 26: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p046_26.png] view at source ↗
Figure 27
Figure 27. Figure 27: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p047_27.png] view at source ↗
Figure 28
Figure 28. Figure 28: Maximum spanning trees of Ridge model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p048_28.png] view at source ↗
Figure 29
Figure 29. Figure 29: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p049_29.png] view at source ↗
Figure 30
Figure 30. Figure 30: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p050_30.png] view at source ↗
Figure 31
Figure 31. Figure 31: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p051_31.png] view at source ↗
Figure 32
Figure 32. Figure 32: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p052_32.png] view at source ↗
Figure 33
Figure 33. Figure 33: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p053_33.png] view at source ↗
Figure 34
Figure 34. Figure 34: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p054_34.png] view at source ↗
Figure 35
Figure 35. Figure 35: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p055_35.png] view at source ↗
Figure 36
Figure 36. Figure 36: Maximum spanning trees of SVR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p056_36.png] view at source ↗
Figure 37
Figure 37. Figure 37: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p057_37.png] view at source ↗
Figure 38
Figure 38. Figure 38: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p058_38.png] view at source ↗
Figure 39
Figure 39. Figure 39: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p059_39.png] view at source ↗
Figure 40
Figure 40. Figure 40: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p060_40.png] view at source ↗
Figure 41
Figure 41. Figure 41: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p061_41.png] view at source ↗
Figure 42
Figure 42. Figure 42: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p062_42.png] view at source ↗
Figure 43
Figure 43. Figure 43: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p063_43.png] view at source ↗
Figure 44
Figure 44. Figure 44: Maximum spanning trees of RFR model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p064_44.png] view at source ↗
Figure 45
Figure 45. Figure 45: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p065_45.png] view at source ↗
Figure 46
Figure 46. Figure 46: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p066_46.png] view at source ↗
Figure 47
Figure 47. Figure 47: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p067_47.png] view at source ↗
Figure 48
Figure 48. Figure 48: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p068_48.png] view at source ↗
Figure 49
Figure 49. Figure 49: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p069_49.png] view at source ↗
Figure 50
Figure 50. Figure 50: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p070_50.png] view at source ↗
Figure 51
Figure 51. Figure 51: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p071_51.png] view at source ↗
Figure 52
Figure 52. Figure 52: Maximum spanning trees of Lasso model feature importance matrices for the emotion [PITH_FULL_IMAGE:figures/full_fig_p072_52.png] view at source ↗
read the original abstract

Decoding emotional states from neural signals has been typically framed as a discrete, single-label classification task based on emotionally stable stimuli, a formulation that oversimplifies the continuous, fluid, and co-occurring nature of human affect. This study reconceptualizes emotion decoding by adopting a multi-target regression framework to track multiple overlapping emotional dimensions as continuous trajectories over time. Leveraging the robust generalization capabilities of Large Language Models (LLMs), we extracted fine-grained, continuous sentiment profiles from a naturalistic auditory narrative, Alice in Wonderland, to serve as scalable proxies for subjective affect from human fMRI dataset. Departing from standard classification paradigms or mass-univariate subtractive contrasts that filter out network dynamics, we leverage regularized and kernel-based machine learning algorithms as continuous estimators to track the magnitude of macroscale neural state variations. We demonstrate that models trained on temporal snapshots of Dynamic Functional Connectivity (DFC) significantly outperform static region-of-interest (ROI) amplitude representations, effectively capturing continuous emotional trajectories under rapidly fluctuating narrative input. Furthermore, by implementing graph-theoretical Explainable AI (XAI) techniques, we deconstruct the underlying predictive features to reveal highly interpretable, emotion-specific topological configurations. Collectively, these results highlight the utility of LLM-automated annotation in affective neuroscience and provide compelling empirical evidence for psychological constructionist frameworks, demonstrating that dynamic, distributed network interactions offer superior explanatory power over strictly locationist accounts of emotion.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes an LLM-enhanced multi-target regression framework to decode continuous, multi-dimensional emotional trajectories from fMRI data acquired during naturalistic narrative listening (Alice in Wonderland). LLM-derived fine-grained sentiment profiles serve as regression targets; regularized and kernel-based models trained on temporal snapshots of dynamic functional connectivity (DFC) are claimed to significantly outperform those using static ROI amplitude features; graph-theoretical XAI is used to recover emotion-specific network topologies, supporting constructionist over locationist accounts of emotion.

Significance. If the LLM-derived targets are validated against human affective measures and the reported DFC superiority is confirmed with appropriate statistics, the work would supply a scalable annotation method for naturalistic stimuli and direct empirical support for dynamic network models of emotion. The combination of continuous regression, DFC, and topological XAI is a coherent methodological contribution to affective decoding.

major comments (2)
  1. [Abstract] Abstract: The central claim that DFC models capture continuous emotional trajectories rests on LLM-extracted sentiment profiles serving as faithful proxies for participants' subjective affect. No correlation with self-reports, behavioral ratings, or physiological validation on the same stimulus is described, leaving open the possibility that the targets primarily reflect textual narrative features rather than individual affective experience; this directly undermines interpretability of the DFC-vs-ROI comparison for the stated psychological conclusions.
  2. [Abstract] Abstract: The assertion that DFC models 'significantly outperform' static ROI representations is presented without any quantitative metrics, cross-validation procedure, statistical tests, effect sizes, or error bars. Because this outperformance is the primary empirical result supporting the framework's superiority, its absence prevents evaluation of whether the result is load-bearing or merely suggestive.
minor comments (2)
  1. [Abstract] The abstract refers to 'regularized and kernel-based machine learning algorithms' without naming the specific estimators (e.g., ridge, SVR, kernel ridge) or regularization parameters; this should be stated explicitly in the methods section for reproducibility.
  2. [Abstract] Dataset details (number of participants, fMRI acquisition parameters, preprocessing pipeline) are not summarized even at a high level in the abstract, which is standard for neuroimaging studies.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the detailed and constructive feedback. We address the major comments point-by-point below, agreeing where revisions are needed to improve clarity and acknowledging limitations where data is unavailable.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that DFC models capture continuous emotional trajectories rests on LLM-extracted sentiment profiles serving as faithful proxies for participants' subjective affect. No correlation with self-reports, behavioral ratings, or physiological validation on the same stimulus is described, leaving open the possibility that the targets primarily reflect textual narrative features rather than individual affective experience; this directly undermines interpretability of the DFC-vs-ROI comparison for the stated psychological conclusions.

    Authors: We acknowledge that the manuscript does not include direct validation of the LLM-derived sentiment profiles against participant self-reports or physiological measures from the same dataset, as the publicly available Alice in Wonderland fMRI dataset does not provide such concurrent affective ratings. Our strongest defense is that LLMs have been shown in prior literature to produce sentiment profiles that correlate highly with human judgments on narrative texts, making them reasonable scalable proxies. We will revise the abstract and add a dedicated limitations paragraph in the discussion to explicitly frame the targets as proxies and discuss this caveat, thereby strengthening the interpretability section without overstating the claims. revision: yes

  2. Referee: [Abstract] Abstract: The assertion that DFC models 'significantly outperform' static ROI representations is presented without any quantitative metrics, cross-validation procedure, statistical tests, effect sizes, or error bars. Because this outperformance is the primary empirical result supporting the framework's superiority, its absence prevents evaluation of whether the result is load-bearing or merely suggestive.

    Authors: The full manuscript reports these details in the Results section, including performance metrics (e.g., R^2 values), cross-validation procedures (leave-one-subject-out), statistical tests (permutation tests), effect sizes, and error bars in figures and tables. However, we agree that the abstract should be more informative. We will revise the abstract to include key quantitative results, such as the reported outperformance margins and significance levels, to allow readers to evaluate the strength of the findings directly from the abstract. revision: yes

standing simulated objections not resolved
  • Direct correlations between LLM sentiment profiles and participant self-reports or physiological data on the Alice in Wonderland stimulus cannot be provided, as such measures are not available in the dataset used.

Circularity Check

0 steps flagged

No significant circularity; regression targets and performance metrics are independent of brain data inputs

full rationale

The paper trains regression models on DFC and ROI features to predict LLM-derived sentiment trajectories extracted from the narrative text. These targets are generated externally from the stimulus audio transcript and are not derived from or fitted to the fMRI data itself. No equations, self-citations, or uniqueness claims are shown that would reduce the reported DFC superiority to a definitional or fitted-input artifact. The framework is self-contained against external benchmarks (LLM outputs and fMRI recordings) with no load-bearing self-referential steps.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim depends on treating LLM sentiment output as a faithful proxy for human affect and on the validity of DFC snapshots as sufficient input features; no free parameters or invented entities are explicitly introduced in the abstract.

axioms (1)
  • domain assumption LLM-extracted sentiment profiles serve as valid proxies for subjective human affect during narrative listening
    Invoked when the paper states these profiles serve as scalable proxies for the fMRI dataset without reported human validation.

pith-pipeline@v0.9.1-grok · 5820 in / 1124 out tokens · 20928 ms · 2026-06-27T22:26:50.871910+00:00 · methodology

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

127 extracted references · 18 canonical work pages

  1. [1]

    bioRxiv , pages=

    GPT-4 accurately predicts human emotions and their neural correlates , author=. bioRxiv , pages=. 2025 , publisher=

    2025

  2. [2]

    Neurocomputing , pages=

    Decoding neural emotion patterns through large language model embeddings , author=. Neurocomputing , pages=. 2025 , publisher=

    2025

  3. [3]

    Frontiers in Neuroscience , volume=

    Explainable AI: A review of applications to neuroimaging data , author=. Frontiers in Neuroscience , volume=. 2022 , publisher=

    2022

  4. [4]

    World Conference on Explainable Artificial Intelligence , pages=

    Explainable emotion decoding for human and computer vision , author=. World Conference on Explainable Artificial Intelligence , pages=. 2024 , organization=

    2024

  5. [5]

    and Davis, W

    Murie, A. and Davis, W. B. The Wolves of Mount McKinley. Fauna of the National Parks of the United States. Journal of Mammalogy. 1945. doi:https://doi.org/10.2307/1375039

    work page doi:10.2307/1375039 1945

  6. [6]

    Mitchell, T. M. Machine Learning. 1997

    1997

  7. [7]

    An Impossibility Theorem for Clustering

    Kleinberg, J. An Impossibility Theorem for Clustering. Proceedings of the 15th International Conference on Neural Information Processing Systems. 2002

    2002

  8. [8]

    Doughty, A. K. and Horton, B. J. and Nguyen, T.D. and Huyen, C. and Ballagh, R. and Corkrey, R. and Hinch, G. N. The influence of lameness and individuality on movement patterns in sheep. Elsevier. 2018

    2018

  9. [9]

    arXiv preprint arXiv:1706.03762 , volume=

    Attention Is All You Need.(Nips), 2017 , author=. arXiv preprint arXiv:1706.03762 , volume=

    Pith/arXiv arXiv 2017

  10. [10]

    Emotion Review , year =

    Barrett, Lisa Feldman , title =. Emotion Review , year =

  11. [11]

    , title =

    Barrett, Lisa Feldman, Tugade, Michele M., and Engle, Randall W. , title =. Psychological Bulletin , year =

  12. [12]

    Perspectives on Psychological Science , year =

    Barrett, Lisa Feldman , title =. Perspectives on Psychological Science , year =

  13. [13]

    Journal of Cognitive Neuroscience , year =

    Barrett, Lisa Feldman and Bliss-Moreau, Eliza , title =. Journal of Cognitive Neuroscience , year =

  14. [14]

    and Barsalou, Lawrence W

    Barrett, Lisa Feldman and Wilson-Mendenhall, Christine D. and Barsalou, Lawrence W. , title =. The Psychological Construction of Emotion , editor =. 2015 , pages =

    2015

  15. [15]

    Social Cognitive and Affective Neuroscience , volume =

    Lisa Feldman Barrett , title =. Social Cognitive and Affective Neuroscience , volume =. 2017 , doi =

    2017

  16. [16]

    , author=

    Neural networks and physical systems with emergent collective computational abilities. , author=. Proceedings of the National Academy of Sciences of the United States of America , year=

  17. [17]

    arXiv preprint arXiv:2001.08361 , year=

    Scaling laws for neural language models , author=. arXiv preprint arXiv:2001.08361 , year=

    Pith/arXiv arXiv 2001

  18. [18]

    arXiv preprint arXiv:2005.14165 , year=

    Language models are few-shot learners , author=. arXiv preprint arXiv:2005.14165 , year=

    Pith/arXiv arXiv 2005

  19. [19]

    2018 , url=

    Improving Language Understanding by Generative Pre-Training , author=. 2018 , url=

    2018

  20. [20]

    2019 , url=

    Language Models are Unsupervised Multitask Learners , author=. 2019 , url=

    2019

  21. [21]

    1980 , publisher =

    Robert Plutchik , title =. 1980 , publisher =

    1980

  22. [22]

    Advances in Neural Information Processing Systems (NeurIPS) , year =

    Jon Gauthier and Roger Levy , title =. Advances in Neural Information Processing Systems (NeurIPS) , year =

  23. [23]

    Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

    Yichen Chen and others , title =. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , year =

  24. [24]

    Machine Learning , volume=

    Bagging Predictors , author=. Machine Learning , volume=. 1996 , publisher=

    1996

  25. [25]

    RTI Press publication No

    Functional Magnetic Resonance Imaging (fMRI): An Invaluable Tool in Translational Neuroscience , author=. RTI Press publication No. OP-0010-1212. Research Triangle Park, NC: RTI Press , year=

  26. [26]

    Neurosurg Clin N Am

    Overview of Functional Magnetic Resonance Imaging , author=. Neurosurg Clin N Am. 2011 April ; 22(2): 133–139. doi:10.1016/j.nec.2010.11.001. , year=

    work page doi:10.1016/j.nec.2010.11.001 2011

  27. [27]

    Human Brain Mapping , volume =

    Riitta Salmelin and Sylvain Baillet , title =. Human Brain Mapping , volume =. 2009 , doi =

    2009

  28. [28]

    Swaab and K

    T.Y. Swaab and K. Ledoux and C.C. Camblin and M.A. Boudewyn , title =. The Oxford Handbook of Event-Related Potential Components , editor =. 2012 , publisher =

    2012

  29. [29]

    IMT Software Wiki - LaTeX

    NTNU, Department of Marine Technology. IMT Software Wiki - LaTeX

  30. [30]

    GigaScience , year =

    Michael Hanke and others , title =. GigaScience , year =

  31. [31]

    Brennan and Wen-Ming Luh and Berta Franzluebbers and John T

    Shohini Bhattasali and Jonathan R. Brennan and Wen-Ming Luh and Berta Franzluebbers and John T. Hale , title =. 2020 , doi =

    2020

  32. [32]

    Cerebral Cortex , year =

    Alexander Schaefer and others , title =. Cerebral Cortex , year =

  33. [33]

    Gamage and D

    G. Gamage and D. De Silva and N. Mills and D. Alahakoon , title =. Journal of Big Data , year =

  34. [34]

    Journal of cognitive neuroscience , volume=

    Face-specific processing in the human fusiform gyrus , author=. Journal of cognitive neuroscience , volume=. 1997 , publisher=

    1997

  35. [35]

    Annual Review of Linguistics , volume=

    Neurocomputational models of language processing , author=. Annual Review of Linguistics , volume=. 2022 , publisher=

    2022

  36. [36]

    Authorea Preprints , year =

    MU Hadi and Q Al Tashi and A Shah and R Qureshi , title =. Authorea Preprints , year =. doi:10.36227/techrxiv.23589741.v6 , url =

    work page doi:10.36227/techrxiv.23589741.v6

  37. [37]

    and Glerean, E

    Saarimäki, H. and Glerean, E. and Smirnov, D. and Mynttinen, H. and Jääskeläinen, I.P. and Sams, M. and Nummenmaa, L. , title =. NeuroImage , volume =. 2022 , doi =

    2022

  38. [38]

    and Zhang, Z

    Xu, S. and Zhang, Z. and Li, L. and Zhou, Y. and Lin, D. and Zhang, M. and Zhang, L. and Huang, G. and Liu, X. and Becker, B. and others , title =. NeuroImage , volume =. 2023 , doi =

    2023

  39. [39]

    2024 , eprint=

    GPT-4 Technical Report , author=. 2024 , eprint=

    2024

  40. [40]

    Scientific reports , volume=

    EEG-based functional and effective connectivity patterns during emotional episodes using graph theoretical analysis , author=. Scientific reports , volume=. 2025 , publisher=

    2025

  41. [41]

    Identifying the hierarchical emotional areas in the human brain through information fusion , journal =

    Zhongyu Huang and Changde Du and Chaozhuo Li and Kaicheng Fu and Huiguang He , keywords =. Identifying the hierarchical emotional areas in the human brain through information fusion , journal =. 2025 , issn =. doi:https://doi.org/10.1016/j.inffus.2024.102613 , url =

    work page doi:10.1016/j.inffus.2024.102613 2025

  42. [42]

    Science China Life Sciences , volume=

    Decoding six basic emotions for brain functional connectivity patterns , author=. Science China Life Sciences , volume=. 2023 , publisher=

    2023

  43. [43]

    Biomedical Signal Processing and Control , volume=

    Exploring emotional experiences and dataset construction in the era of short videos based on physiological signals , author=. Biomedical Signal Processing and Control , volume=. 2024 , publisher=

    2024

  44. [44]

    Neuropsychologia , volume=

    The brain basis of emotion: A meta-analytic review , author=. Neuropsychologia , volume=. 2011 , publisher=

    2011

  45. [45]

    Computational and Mathematical Methods in Medicine , year =

    Mahmoudi, Abdelhak and Takerkart, Sylvain and Regragui, Fakhita and Boussaoud, Driss and Brovelli, Andrea , title =. Computational and Mathematical Methods in Medicine , year =. doi:10.1155/2012/961257 , pmcid =

    work page doi:10.1155/2012/961257 2012

  46. [46]

    Pattern Recognition Letters , volume=

    Identifying the hierarchical emotional areas in the human brain through information fusion , author=. Pattern Recognition Letters , volume=. 2024 , publisher=

    2024

  47. [47]

    Pattern Recognition Letters , volume=

    Multimodal Emotion Recognition with Deep Learning: Advancements, challenges, and future directions , author=. Pattern Recognition Letters , volume=. 2023 , publisher=

    2023

  48. [48]

    Neurocomputing , volume=

    Emotion Detection from Multimodal Data Using Machine Learning , author=. Neurocomputing , volume=. 2020 , doi=

    2020

  49. [49]

    Psychological Review , volume=

    Core Affect and the Psychological Construction of Emotion , author=. Psychological Review , volume=. 2003 , doi=

    2003

  50. [50]

    MIT Press , year=

    Positive Computing: Technology for Wellbeing and Human Potential , author=. MIT Press , year=

  51. [51]

    Current Directions in Psychological Science , volume=

    The Nature of Pride , author=. Current Directions in Psychological Science , volume=. 2007 , doi=

    2007

  52. [52]

    Current Biology , volume =

    The Human Amygdala and the Induction and Experience of Fear , author =. Current Biology , volume =. 2011 , publisher =. doi:10.1016/j.cub.2010.11.042 , pmid =

    work page doi:10.1016/j.cub.2010.11.042 2011

  53. [53]

    NeuroImage , volume=

    Functional Neuroanatomy of Emotion: A Meta-Analysis of Emotion Activation Studies in PET and fMRI , author=. NeuroImage , volume=. 2002 , doi=

    2002

  54. [54]

    NeuroImage , year =

    Functional grouping and cortical-subcortical interactions in emotion: a meta-analysis of neuroimaging studies , author =. NeuroImage , year =. doi:10.1016/j.neuroimage.2008.03.059 , pmid =

    work page doi:10.1016/j.neuroimage.2008.03.059 2008

  55. [55]

    Nature Neuroscience , volume=

    Neurobiology of Anxiety: From Neural Circuits to Novel Solutions , author=. Nature Neuroscience , volume=. 2007 , doi=

    2007

  56. [56]

    NeuroImage , volume=

    Dynamic Functional Connectivity: Promise, Issues, and Interpretations , author=. NeuroImage , volume=. 2013 , doi=

    2013

  57. [57]

    LeDoux , keywords =

    Joseph E. LeDoux , keywords =. The amygdala: contributions to fear and stress , journal =. 1994 , issn =. doi:https://doi.org/10.1006/smns.1994.1030 , url =

    work page doi:10.1006/smns.1994.1030 1994

  58. [58]

    Brain Research Reviews , volume=

    Predictors of amygdala activation during the processing of emotional stimuli: a meta-analysis of 385 PET and fMRI studies , author=. Brain Research Reviews , volume=. 2008 , month=. doi:10.1016/j.brainresrev.2007.10.012 , url=

    work page doi:10.1016/j.brainresrev.2007.10.012 2008

  59. [59]

    Satheesh Kumar and P

    J. Satheesh Kumar and P. Bhuvaneswari , keywords =. Analysis of Electroencephalography (EEG) Signals and Its Categorization–A Study , journal =. 2012 , note =. doi:https://doi.org/10.1016/j.proeng.2012.06.298 , url =

    work page doi:10.1016/j.proeng.2012.06.298 2012

  60. [60]

    and Jastorff, J

    Huang, Y.-A. and Jastorff, J. and Van den Stock, J. and Van de Vliet, L. and Dupont, P. and Vandenbulcke, M. , title =. NeuroImage , volume =. 2018 , doi =

    2018

  61. [61]

    Nature Reviews Neuroscience , year =

    Tim Dalgleish , title =. Nature Reviews Neuroscience , year =. doi:10.1038/nrn1432 , url =

    work page doi:10.1038/nrn1432

  62. [62]

    Cognition and Emotion , year =

    Ekman, Paul , title =. Cognition and Emotion , year =

  63. [63]

    Rached and A

    T.S. Rached and A. Perkusich , title =. Brain-Computer Interface Systems , editor =. 2013 , publisher =. doi:10.5772/56227 , url =

    work page doi:10.5772/56227 2013

  64. [64]

    Proceedings of the National Academy of Sciences , volume=

    Neural networks and physical systems with emergent collective computational abilities , author=. Proceedings of the National Academy of Sciences , volume=. 1982 , doi=

    1982

  65. [65]

    and Simard, P

    Bengio, Y. and Simard, P. and Frasconi, P. , journal=. Learning long-term dependencies with gradient descent is difficult , year=

  66. [66]

    Neural Computation , volume=

    Long Short-Term Memory , author=. Neural Computation , volume=. 1997 , doi=

    1997

  67. [67]

    Proceedings of the IEEE , volume=

    Gradient-based learning applied to document recognition , author=. Proceedings of the IEEE , volume=. 1998 , doi=

    1998

  68. [68]

    Kipf and Max Welling , title =

    Thomas N. Kipf and Max Welling , title =. arXiv preprint arXiv:1609.02907 , year =

    Pith/arXiv arXiv

  69. [69]

    right to explanation

    European Union regulations on algorithmic decision-making and a “right to explanation” , author=. AI magazine , volume=

  70. [70]

    The Lancet Digital Health , volume=

    The false hope of current approaches to explainable artificial intelligence in health care , author=. The Lancet Digital Health , volume=. 2021 , publisher=

    2021

  71. [71]

    2016 , publisher=

    Deep Learning , author=. 2016 , publisher=

    2016

  72. [72]

    Cerebral Cortex , volume=

    Discrete Neural Signatures of Basic Emotions , author=. Cerebral Cortex , volume=. 2016 , publisher=

    2016

  73. [73]

    PLOS ONE , volume=

    Identifying Emotions on the Basis of Neural Activation , author=. PLOS ONE , volume=. 2013 , publisher=

    2013

  74. [74]

    PLoS Computational Biology , volume=

    A Bayesian Model of Category-Specific Emotional Brain Responses , author=. PLoS Computational Biology , volume=. 2015 , publisher=

    2015

  75. [75]

    Brown and Steven P

    Jeremy Kawahara and Colin J. Brown and Steven P. Miller and Brian G. Booth and Vann Chau and Ruth E. Grunau and Jill G. Zwicker and Ghassan Hamarneh , keywords =. BrainNetCNN: Convolutional neural networks for brain networks; towards predicting neurodevelopment , journal =. 2017 , issn =. doi:https://doi.org/10.1016/j.neuroimage.2016.09.046 , url =

    work page doi:10.1016/j.neuroimage.2016.09.046 2017

  76. [76]

    2022 , eprint=

    Brain Network Transformer , author=. 2022 , eprint=

    2022

  77. [77]

    IEEE Transactions on Medical Imaging , volume=

    Graph-Enhanced Emotion Neural Decoding , author=. IEEE Transactions on Medical Imaging , volume=. 2023 , month=. doi:10.1109/TMI.2023.3246220 , publisher=

    work page doi:10.1109/tmi.2023.3246220 2023

  78. [78]

    arXiv preprint arXiv:2408.13718 , year=

    GPT-4 Emulates Average-Human Emotional Cognition from a Third-Person Perspective , author=. arXiv preprint arXiv:2408.13718 , year=

    arXiv

  79. [79]

    2019 , howpublished=

    Narratives: fMRI data for evaluating models of naturalistic language comprehension , author=. 2019 , howpublished=

    2019

  80. [80]

    Scientific Data , volume=

    A high-resolution 7-Tesla fMRI dataset from complex natural stimulation with an audio movie , author=. Scientific Data , volume=. 2014 , publisher=. doi:10.1038/sdata.2014.3 , url=

    work page doi:10.1038/sdata.2014.3 2014

Showing first 80 references.