pith. sign in

arxiv: 2605.20389 · v1 · pith:CAUDPD6Anew · submitted 2026-05-19 · 💻 cs.LG · cs.AI

Nonlocal operator learning for fMRI encoding and decoding tasks

Andreas Kramer , Saugat Acharya , Alice Giola , Emanuele Zappala This is my paper

Pith reviewed 2026-05-21 08:05 UTC · model grok-4.3

classification 💻 cs.LG cs.AI
keywords fMRIneural integral operatorsspatiotemporal contextencoding and decodinglatent representationsnonlocal dependenciesbrain dynamicsfunctional MRI
0
0 comments X

The pith

Neural integral operators model fMRI dynamics by capturing nonlocal spatiotemporal context through latent fixed-point iterations.

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

The paper tests whether neural integral operator models can handle the complex high-dimensional patterns in fMRI recordings for both decoding stimuli from brain activity and encoding brain responses from stimulus input. It centers on the value of nonlocal context by comparing short versus long temporal windows and visual cortex versus whole-brain data across two open datasets. Results indicate that longer windows tend to raise performance and yield more organized latent representations, while the learned space often separates stimulus classes more clearly than the original recordings do. A sympathetic reader would care because fMRI data is notoriously noisy and distributed, so architectures built for nonlocal dependencies could support better prediction of brain states or more accurate decoding of mental content.

Core claim

The authors establish that a latent neural integral operator framework, which performs fixed-point iterations in an auxiliary space before classification or prediction, provides a workable approach for fMRI encoding and decoding tasks, and that expanding the spatiotemporal context through longer temporal windows and whole-brain recordings produces measurable gains in accuracy together with more structured latent geometry compared with shorter or more local inputs.

What carries the argument

Latent neural integral operator that performs fixed-point iterations in an auxiliary space to capture nonlocal spatiotemporal dependencies.

If this is right

  • Larger temporal windows produce consistent performance gains in both decoding and encoding across the tested datasets.
  • Whole-brain recordings support better results than visual-cortex-only inputs for these operator-based models.
  • The learned latent space frequently shows sharper separation between stimulus classes than the raw fMRI data in decoding experiments.
  • Longer contexts lead to more structured geometry in the auxiliary latent representations.

Where Pith is reading between the lines

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

  • The same nonlocal operator style could be tested on other distributed neural recordings such as EEG or intracranial signals to check whether the context benefit generalizes.
  • If the latent iterations remain stable at scale, the approach might support real-time decoding systems that integrate information across longer histories than current local models allow.
  • Architectures that explicitly iterate in auxiliary space may offer a route to reduce the sample complexity of learning distributed brain dynamics compared with purely local convolutional or recurrent networks.

Load-bearing premise

Fixed-point iterations inside the auxiliary latent space can reliably locate and use the nonlocal spatiotemporal dependencies present in fMRI recordings.

What would settle it

Running the same decoding and encoding tasks on the same datasets but with a non-integral baseline model or with fixed short windows and finding equal or higher accuracy plus equally structured representations would indicate the integral operator and broader context add no advantage.

Figures

Figures reproduced from arXiv: 2605.20389 by Alice Giola, Andreas Kramer, Emanuele Zappala, Saugat Acharya.

Figure 1
Figure 1. Figure 1: Schematic description of fMRI data acquisition and preprocessing. Panel 1 [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Schematic description of decoding and encoding preoblems. In the decoding [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Example embedding of the latent space created by the neural integral operator. [PITH_FULL_IMAGE:figures/full_fig_p012_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Example embedding of the raw data. The corresponding KNN classification [PITH_FULL_IMAGE:figures/full_fig_p013_4.png] view at source ↗
read the original abstract

Functional MRI data exhibit high-dimensional spatiotemporal structure, making both prediction and decoding challenging. In this work, we investigate neural integral-operator-based models for encoding and decoding tasks in fMRI, with particular emphasis on the role of nonlocal spatiotemporal context. We implement a latent neural integral operator framework that performs fixed point iterations in an auxiliary space from which classification and stimuli prediction is performed via a decoder. We evaluate our model on two open-source fMRI datasets. Our experiments examine both decoding of stimuli from fMRI recordings and encoding of fMRI dynamics from stimulus representations. A main focus is the effect of spatiotemporal context: we systematically compare short and long temporal windows, as well as the use of visual cortex vs whole brain recordings, and analyze their influence on performance and latent-space geometry. Across tasks and datasets, larger temporal windows generally improve results and produce more structured learned representations. In decoding experiments, the learned latent space often provides clearer class separation than the raw data. In encoding experiments, although absolute performance remains moderate due to the difficulty of the task, longer temporal windows still yield consistent gains. These findings suggest that neural integral operators provide a promising framework for modeling fMRI dynamics and that broader spatiotemporal context can be beneficial for both prediction and representation learning. More broadly, the results indicate that exploiting distributed nonlocal structure in brain dynamics requires model architectures specifically designed to capture such dependencies.

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

3 major / 2 minor

Summary. The manuscript proposes a latent neural integral operator framework for fMRI encoding and decoding tasks. Fixed-point iterations are performed in an auxiliary latent space to capture nonlocal spatiotemporal dependencies, after which a decoder handles classification or stimulus prediction. Experiments on two open fMRI datasets systematically vary temporal window length and input region (visual cortex versus whole brain), reporting that longer windows improve performance and yield more structured latent representations with clearer class separation in decoding tasks.

Significance. If the performance gains can be attributed specifically to the nonlocal operator mechanism rather than raw increases in context length or model capacity, the work would offer a useful architecture for high-dimensional spatiotemporal brain data. The focus on open datasets and controlled variation of spatiotemporal context provides a concrete empirical test bed for operator-based models in neuroimaging.

major comments (3)
  1. Abstract: the claim that 'larger temporal windows generally improve results and produce more structured learned representations' supplies no quantitative metrics, error bars, statistical tests, or description of data splits and evaluation procedures, rendering the central empirical claims unverifiable from the provided text.
  2. Abstract / Model Description: the central modeling choice that fixed-point iterations in the auxiliary latent space extract and exploit nonlocal spatiotemporal dependencies is not supported by ablation studies, convergence analysis of the iterations, or comparisons to capacity-matched baselines (e.g., standard RNN or attention layers on identical windows). Without such controls, observed gains could be explained by increased input context alone.
  3. Experiments section: no direct comparison is reported between the proposed operator model and simpler architectures with equivalent parameter counts or context lengths, which is required to isolate the contribution of the neural integral operator framework to the reported improvements in both decoding and encoding tasks.
minor comments (2)
  1. Abstract: the two open-source fMRI datasets are not named or referenced, which hinders immediate reproducibility.
  2. The description of latent-space geometry and class separation would benefit from explicit quantification (e.g., silhouette scores or visualization details) rather than qualitative statements.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. The comments identify key areas where additional clarity and controls will strengthen the manuscript. We address each major comment below and commit to the indicated revisions.

read point-by-point responses

  1. Referee: Abstract: the claim that 'larger temporal windows generally improve results and produce more structured learned representations' supplies no quantitative metrics, error bars, statistical tests, or description of data splits and evaluation procedures, rendering the central empirical claims unverifiable from the provided text.

    Authors: We agree that the abstract, due to length constraints, presents the findings at a high level without the supporting quantitative details. In the revised manuscript we will augment the abstract with representative performance metrics (including means and standard deviations across runs), and we will ensure the methods and results sections explicitly describe the data splits, cross-validation procedures, and statistical tests used to support the claims regarding temporal window length and latent-space structure. revision: yes

  2. Referee: Abstract / Model Description: the central modeling choice that fixed-point iterations in the auxiliary latent space extract and exploit nonlocal spatiotemporal dependencies is not supported by ablation studies, convergence analysis of the iterations, or comparisons to capacity-matched baselines (e.g., standard RNN or attention layers on identical windows). Without such controls, observed gains could be explained by increased input context alone.

    Authors: We acknowledge that the current manuscript does not contain explicit ablation studies that isolate the contribution of the fixed-point iterations or convergence diagnostics for those iterations. While the architecture is derived from the neural integral operator framework precisely to enable nonlocal interactions via the auxiliary latent space, we will add (i) an ablation that replaces the fixed-point iteration block with a single forward pass of equivalent depth, (ii) convergence plots for the iterations across representative samples, and (iii) comparisons against capacity-matched RNN and attention baselines that receive identical temporal windows. These additions will appear in a new subsection of the experiments. revision: yes

  3. Referee: Experiments section: no direct comparison is reported between the proposed operator model and simpler architectures with equivalent parameter counts or context lengths, which is required to isolate the contribution of the neural integral operator framework to the reported improvements in both decoding and encoding tasks.

    Authors: We agree that matched-parameter and matched-context comparisons are necessary to attribute gains specifically to the operator-based mechanism rather than to increased context or capacity. In the revised version we will report results for the proposed model alongside RNN, LSTM, and transformer baselines that are configured to have approximately the same number of parameters and that operate on the same temporal windows and spatial regions. These comparisons will be presented for both the decoding and encoding tasks on the two datasets. revision: yes

Circularity Check

0 steps flagged

No significant circularity; empirical evaluation on open datasets

full rationale

The paper presents an empirical study implementing a latent neural integral operator framework for fMRI tasks and evaluating it on open-source datasets via comparisons of temporal windows and brain regions. No mathematical derivation chain exists that reduces predictions or results to inputs by construction, self-definition, or self-citation load-bearing. Performance gains and latent-space observations are reported from direct experiments rather than fitted quantities renamed as predictions. The fixed-point iteration modeling choice is an architectural decision tested empirically, with no uniqueness theorems or ansatzes imported circularly from prior self-work.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, background axioms, or newly postulated entities are stated in the provided text. The framework presumably contains learned weights inside the integral operator and decoder, but these cannot be enumerated or classified without the full manuscript.

pith-pipeline@v0.9.0 · 5777 in / 1177 out tokens · 52954 ms · 2026-05-21T08:05:35.499765+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

41 extracted references · 41 canonical work pages

  1. [1]

    Decoding hand kinematics from local field potentials using long short-term memory (lstm) network

    Nur Ahmadi, Timothy G Constandinou, and Christos-Savvas Bouganis. Decoding hand kinematics from local field potentials using long short-term memory (lstm) network. In 2019 9th international IEEE/EMBS conference on neural engineering (NER) , pages 415--419. IEEE, 2019

    work page 2019

  2. [2]

    Homogeneous nets of neuron-like elements

    Shun-Ichi Amari. Homogeneous nets of neuron-like elements. Biological cybernetics , 17(4):211--220, 1975

    work page 1975

  3. [3]

    Dynamics of pattern formation in lateral-inhibition type neural fields

    Shun-ichi Amari. Dynamics of pattern formation in lateral-inhibition type neural fields. Biological cybernetics , 27(2):77--87, 1977

    work page 1977

  4. [4]

    A transformer model for learning spatiotemporal contextual representation in fmri data

    Nima Asadi, Ingrid R Olson, and Zoran Obradovic. A transformer model for learning spatiotemporal contextual representation in fmri data. Network Neuroscience , 7(1):22--47, 2023

    work page 2023

  5. [5]

    Bolt: Fused window transformers for fmri time series analysis

    Hasan A Bedel, Irmak Sivgin, Onat Dalmaz, Salman UH Dar, and Tolga C ukur. Bolt: Fused window transformers for fmri time series analysis. Medical image analysis , 88:102841, 2023

    work page 2023

  6. [6]

    Wilson--cowan equations for neocortical dynamics

    Jack D Cowan, Jeremy Neuman, and Wim van Drongelen. Wilson--cowan equations for neocortical dynamics. The Journal of Mathematical Neuroscience , 6(1):1, 2016

    work page 2016

  7. [7]

    Spatiotemporal characterization of the functional mri latency structure with respect to neural signaling and brain hierarchy

    Hyoungshin Choi, Yeongjun Park, Jong-eun Lee, Sunghun Kim, Bo-yong Park, and Hyunjin Park. Spatiotemporal characterization of the functional mri latency structure with respect to neural signaling and brain hierarchy. Advanced Science , 12(43):e04956, 2025

    work page 2025

  8. [8]

    Spatio-temporal correlation tensors reveal functional structure in human brain

    Zhaohua Ding, Allen T Newton, Ran Xu, Adam W Anderson, Victoria L Morgan, and John C Gore. Spatio-temporal correlation tensors reveal functional structure in human brain. PloS one , 8(12):e82107, 2013

    work page 2013

  9. [9]

    The human brain is intrinsically organized into dynamic, anticorrelated functional networks

    Michael D Fox, Abraham Z Snyder, Justin L Vincent, Maurizio Corbetta, David C Van Essen, and Marcus E Raichle. The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proceedings of the National Academy of Sciences , 102(27):9673--9678, 2005

    work page 2005

  10. [10]

    Machine learning for neural decoding

    Joshua I Glaser, Ari S Benjamin, Raeed H Chowdhury, Matthew G Perich, Lee E Miller, and Konrad P Kording. Machine learning for neural decoding. eneuro , 7(4), 2020

    work page 2020

  11. [11]

    3d masked autoencoder with spatiotemporal transformer for modeling of 4d fmri data

    Jie Gao, Bao Ge, Ning Qiang, and Shijie Zhao. 3d masked autoencoder with spatiotemporal transformer for modeling of 4d fmri data. Medical Image Analysis , page 103861, 2025

    work page 2025

  12. [12]

    Deep learning models reveal the link between dynamic brain connectivity patterns and states of consciousness

    Chlo \'e Gomez, Lynn Uhrig, Vincent Frouin, Edouard Duchesnay, B \'e chir Jarraya, and Antoine Grigis. Deep learning models reveal the link between dynamic brain connectivity patterns and states of consciousness. Scientific Reports , 14(1):31606, 2024

    work page 2024

  13. [13]

    Scale-free properties of the functional magnetic resonance imaging signal during rest and task

    Biyu J He. Scale-free properties of the functional magnetic resonance imaging signal during rest and task. Journal of Neuroscience , 31(39):13786--13795, 2011

    work page 2011

  14. [14]

    Distributed and overlapping representations of faces and objects in ventral temporal cortex

    James V Haxby, M Ida Gobbini, Maura L Furey, Alumit Ishai, Jennifer L Schouten, and Pietro Pietrini. Distributed and overlapping representations of faces and objects in ventral temporal cortex. Science , 293(5539):2425--2430, 2001

    work page 2001

  15. [15]

    Rethinking functional brain connectome analysis: Do graph deep learning models help?, 2025

    Keqi Han, Yao Su, Lifang He, Liang Zhan, Sergey Plis, Vince Calhoun, and Carl Yang. Rethinking functional brain connectome analysis: Do graph deep learning models help?, 2025. URL https://arxiv. org/abs/2501.17207

    work page arXiv 2025

  16. [16]

    It’s about time: Linking dynamical systems with human neuroimaging to understand the brain

    Yohan J John, Kayle S Sawyer, Karthik Srinivasan, Eli J M \"u ller, Brandon R Munn, and James M Shine. It’s about time: Linking dynamical systems with human neuroimaging to understand the brain. Network Neuroscience , 6(4):960--979, 2022

    work page 2022

  17. [17]

    Swift: Swin 4d fmri transformer

    Peter Kim, Junbeom Kwon, Sunghwan Joo, Sangyoon Bae, Donggyu Lee, Yoonho Jung, Shinjae Yoo, Jiook Cha, and Taesup Moon. Swift: Swin 4d fmri transformer. Advances in Neural Information Processing Systems , 36:42015--42037, 2023

    work page 2023

  18. [18]

    Mental image reconstruction from human brain activity: Neural decoding of mental imagery via deep neural network-based bayesian estimation

    Naoko Koide-Majima, Shinji Nishimoto, and Kei Majima. Mental image reconstruction from human brain activity: Neural decoding of mental imagery via deep neural network-based bayesian estimation. Neural Networks , 170:349--363, 2024

    work page 2024

  19. [19]

    Visual image reconstruction from brain activity via latent representation

    Yukiyasu Kamitani, Misato Tanaka, and Ken Shirakawa. Visual image reconstruction from brain activity via latent representation. Annual Review of Vision Science , 11(2025):611--634, 2025

    work page 2025

  20. [20]

    Consciousness-specific dynamic interactions of brain integration and functional diversity

    Andrea I Luppi, Michael M Craig, Ioannis Pappas, Paola Finoia, Guy B Williams, Judith Allanson, John D Pickard, Adrian M Owen, Lorina Naci, David K Menon, et al. Consciousness-specific dynamic interactions of brain integration and functional diversity. Nature communications , 10(1):4616, 2019

    work page 2019

  21. [21]

    Generative adversarial networks in brain imaging: A narrative review

    Maria Elena Laino, Pierandrea Cancian, Letterio Salvatore Politi, Matteo Giovanni Della Porta, Luca Saba, and Victor Savevski. Generative adversarial networks in brain imaging: A narrative review. Journal of imaging , 8(4):83, 2022

    work page 2022

  22. [22]

    Hongming Li and Yong Fan. Interpretable, highly accurate brain decoding of subtly distinct brain states from functional mri using intrinsic functional networks and long short-term memory recurrent neural networks. NeuroImage , 202:116059, 2019

    work page 2019

  23. [23]

    From descriptive connectome to mechanistic connectome: Generative modeling in functional magnetic resonance imaging analysis

    Guoshi Li and Pew-Thian Yap. From descriptive connectome to mechanistic connectome: Generative modeling in functional magnetic resonance imaging analysis. Frontiers in Human Neuroscience , 16:940842, 2022

    work page 2022

  24. [24]

    Neural decoding and feature selection methods for closed-loop control of avoidance behavior

    Jinhan Liu, Rebecca Younk, Lauren M Drahos, Sumedh S Nagrale, Shreya Yadav, Alik S Widge, and Mahsa Shoaran. Neural decoding and feature selection methods for closed-loop control of avoidance behavior. Journal of neural engineering , 21(5):056041, 2024

    work page 2024

  25. [25]

    Deep learning applications in fmri--a review work

    Jiangxue Li and Peize Zhao. Deep learning applications in fmri--a review work. In Proceedings of the 2023 13th International Conference on Bioscience, Biochemistry and Bioinformatics , pages 75--80, 2023

    work page 2023

  26. [26]

    Graph neural networks in brain connectivity studies: Methods, challenges, and future directions

    Hamed Mohammadi and Waldemar Karwowski. Graph neural networks in brain connectivity studies: Methods, challenges, and future directions. Brain Sciences , 15(1):17, 2024

    work page 2024

  27. [27]

    Visual image reconstruction from human brain activity using a combination of multiscale local image decoders

    Yoichi Miyawaki, Hajime Uchida, Okito Yamashita, Masa-aki Sato, Yusuke Morito, Hiroki C Tanabe, Norihiro Sadato, and Yukiyasu Kamitani. Visual image reconstruction from human brain activity using a combination of multiscale local image decoders. Neuron , 60(5):915--929, 2008

    work page 2008

  28. [28]

    Nilearn contributors . nilearn. https://github.com/nilearn/nilearn, 2023. Software. DOI: 10.5281/zenodo.8397156

    work page doi:10.5281/zenodo.8397156 2023

  29. [29]

    Inferring single-trial neural population dynamics using sequential auto-encoders

    Chethan Pandarinath, Daniel J O’Shea, Jasmine Collins, Rafal Jozefowicz, Sergey D Stavisky, Jonathan C Kao, Eric M Trautmann, Matthew T Kaufman, Stephen I Ryu, Leigh R Hochberg, et al. Inferring single-trial neural population dynamics using sequential auto-encoders. Nature methods , 15(10):805--815, 2018

    work page 2018

  30. [30]

    Amari model

    Roland Potthast. Amari model. In Encyclopedia of Computational Neuroscience , pages 171--175. Springer, 2022

    work page 2022

  31. [31]

    Interpreting models interpreting brain dynamics

    Md Mahfuzur Rahman, Usman Mahmood, Noah Lewis, Harshvardhan Gazula, Alex Fedorov, Zening Fu, Vince D Calhoun, and Sergey M Plis. Interpreting models interpreting brain dynamics. Scientific reports , 12(1):12023, 2022

    work page 2022

  32. [32]

    A neural network that finds a naturalistic solution for the production of muscle activity

    David Sussillo, Mark M Churchland, Matthew T Kaufman, and Krishna V Shenoy. A neural network that finds a naturalistic solution for the production of muscle activity. Nature neuroscience , 18(7):1025--1033, 2015

    work page 2015

  33. [33]

    Breakdown of long-range temporal dependence in default mode and attention networks during deep sleep

    Enzo Tagliazucchi, Frederic von Wegner, Astrid Morzelewski, Verena Brodbeck, Kolja Jahnke, and Helmut Laufs. Breakdown of long-range temporal dependence in default mode and attention networks during deep sleep. Proceedings of the National Academy of Sciences , 110(38):15419--15424, 2013

    work page 2013

  34. [34]

    Ensemble graph neural network model for classification of major depressive disorder using whole-brain functional connectivity

    Sujitha Venkatapathy, Mikhail Votinov, Lisa Wagels, Sangyun Kim, Munseob Lee, Ute Habel, In-Ho Ra, and Han-Gue Jo. Ensemble graph neural network model for classification of major depressive disorder using whole-brain functional connectivity. Frontiers in psychiatry , 14:1125339, 2023

    work page 2023

  35. [35]

    A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue

    Hugh R Wilson and Jack D Cowan. A mathematical theory of the functional dynamics of cortical and thalamic nervous tissue. Kybernetik , 13(2):55--80, 1973

    work page 1973

  36. [36]

    Decoding and mapping task states of the human brain via deep learning

    Xiaoxiao Wang, Xiao Liang, Zhoufan Jiang, Benedictor A Nguchu, Yawen Zhou, Yanming Wang, Huijuan Wang, Yu Li, Yuying Zhu, Feng Wu, et al. Decoding and mapping task states of the human brain via deep learning. Human brain mapping , 41(6):1505--1519, 2020

    work page 2020

  37. [37]

    The organization of the human cerebral cortex estimated by intrinsic functional connectivity

    BT Thomas Yeo, Fenna M Krienen, Jorge Sepulcre, Mert R Sabuncu, Danial Lashkari, Marisa Hollinshead, Joshua L Roffman, Jordan W Smoller, Lilla Z \"o llei, Jonathan R Polimeni, et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. Journal of neurophysiology , 2011

    work page 2011

  38. [38]

    Time-resolved resting-state brain networks

    Andrew Zalesky, Alex Fornito, Luca Cocchi, Leonardo L Gollo, and Michael Breakspear. Time-resolved resting-state brain networks. Proceedings of the National Academy of Sciences , 111(28):10341--10346, 2014

    work page 2014

  39. [39]

    Learning integral operators via neural integral equations

    Emanuele Zappala, Antonio Henrique de Oliveira Fonseca, Josue Ortega Caro, Andrew Henry Moberly, Michael James Higley, Jessica Cardin, and David van Dijk. Learning integral operators via neural integral equations. Nature Machine Intelligence , 6(9):1046--1062, 2024

    work page 2024

  40. [40]

    Neural integro-differential equations

    Emanuele Zappala, Antonio H de O Fonseca, Andrew H Moberly, Michael J Higley, Chadi Abdallah, Jessica A Cardin, and David van Dijk. Neural integro-differential equations. In Proceedings of the aaai conference on artificial intelligence , volume 37, pages 11104--11112, 2023

    work page 2023

  41. [41]

    3d cnn based automatic diagnosis of attention deficit hyperactivity disorder using functional and structural mri

    Liang Zou, Jiannan Zheng, Chunyan Miao, Martin J Mckeown, and Z Jane Wang. 3d cnn based automatic diagnosis of attention deficit hyperactivity disorder using functional and structural mri. Ieee Access , 5:23626--23636, 2017

    work page 2017