pith. sign in

arxiv: 2410.03057 · v2 · submitted 2024-10-04 · 💻 cs.CE

What Causes Performance Degradation in Cross-Subject EEG Classification?

Yihe Wang , Taida Li , Yujun Yan , Wenzhan Song , Xiang Zhang This is my paper

Pith reviewed 2026-05-23 19:56 UTC · model grok-4.3

classification 💻 cs.CE
keywords EEG classificationcross-subject generalizationinter-subject variabilityshortcut learningperformance degradationmotor imagerybrain disease detection
0
0 comments X

The pith

Cross-subject EEG classification degrades due to inter-subject variability in multi-class tasks and shortcut learning in single-class tasks.

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

The paper conducts controlled experiments to determine the causes of lower performance in cross-subject EEG classification compared to subject-dependent settings. It concludes that inter-subject variability primarily affects multi-class-per-subject tasks such as motor imagery, emotion recognition, and ERP stimulus classification, while shortcut learning based on subject-specific features drives degradation in single-class-per-subject tasks like brain disease detection. Understanding these distinct mechanisms matters because it can inform the development of task-appropriate methods and evaluation standards in EEG-based machine learning applications. The work stresses the need for careful experimental design to distinguish between these factors.

Core claim

Through a series of controlled experiments, the authors show that the performance degradation in cross-subject EEG classification is generally attributable to two factors: inter-subject variability for multi-class-per-subject tasks and shortcut learning for single-class-per-subject tasks.

What carries the argument

Controlled experiments that separate the effects of inter-subject variability from shortcut learning across EEG task categories.

If this is right

  • Task-specific strategies are needed to address the dominant degradation factor in each EEG application.
  • Evaluation protocols must account for the task type to accurately assess model performance.
  • Models for brain disease detection should focus on preventing reliance on subject-specific shortcuts.
  • Domain adaptation techniques may be more relevant for motor imagery and similar multi-class tasks.

Where Pith is reading between the lines

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

  • Similar patterns might appear in other physiological signal classifications beyond EEG.
  • Researchers could develop diagnostic tools to detect which degradation factor is at play in a given dataset.
  • Pretraining approaches might need to differ based on whether the target task is multi-class or single-class per subject.

Load-bearing premise

The controlled experiments isolate inter-subject variability from shortcut learning without confounding influences from how the datasets were built, preprocessed, or which model architecture was used.

What would settle it

A new set of experiments showing that the degradation causes do not align with the multi-class versus single-class task distinction would indicate the claim is incorrect.

Figures

Figures reproduced from arXiv: 2410.03057 by Taida Li, Wenzhan Song, Xiang Zhang, Yihe Wang, Yujun Yan.

Figure 1
Figure 1. Figure 1: Types of MedTS Datasets. S and C denote subject and class, respectively. Medical time series (MedTS) are special￾ized time series data representing continuous recordings of physiological signals from hu￾man subjects Wang et al. (2024b), including EEG, ECG, fNIRS, and PPG signals Badr et al. (2024); Liu et al. (2021); Eastmond et al. (2022); Esgalhado et al. (2021). We propose a taxonomy for MedTS evalua￾ti… view at source ↗
Figure 2
Figure 2. Figure 2: Types of MedTS Evaluation Setups. (a) This diagram shows the two main evaluation setups and their sub-types, (b) This figure adopted from Wang et al. (2024b) shows the differences between the two main setups: subject-dependent and subject-independent. subject’s medical state is fixed or dynamic over time. See detailed explanations and real-world exam￾ples in Section 2.1. Depending on the type of MedTS, var… view at source ↗
read the original abstract

Cross-subject EEG classification typically achieves significantly lower performance than subject-dependent settings. Although this phenomenon has been widely observed in the literature, the underlying causes have not been systematically studied. In this paper, we design a series of controlled experiments to investigate the mechanisms behind the performance drop in cross-subject EEG classification across different EEG tasks. We show that the performance degradation can generally be attributed to two factors: inter-subject variability and shortcut learning. Specifically, multi-class-per-subject EEG classification tasks, such as motor imagery, emotion recognition, and ERP stimulus classification, are mainly affected by inter-subject variability, whereas single-class-per-subject EEG classification tasks, such as brain disease detection, are primarily influenced by shortcut learning based on subject-specific features. These findings provide new insights into the challenges of cross-subject EEG classification and emphasize the importance of appropriate evaluation protocols in EEG research. The code is available at https://github.com/DL4mHealth/EEG-Cross-Subject.

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 paper claims that performance degradation in cross-subject EEG classification arises from two main factors: inter-subject variability (dominant in multi-class-per-subject tasks such as motor imagery, emotion recognition, and ERP stimulus classification) and shortcut learning based on subject-specific features (dominant in single-class-per-subject tasks such as brain disease detection). This is supported by a series of controlled experiments across EEG tasks, with the conclusion that appropriate evaluation protocols are needed. Code is released at a public GitHub repository.

Significance. If the central attribution holds after controlling for confounders, the work provides actionable distinctions between task types that could guide future cross-subject EEG modeling and evaluation design. The public code release is a clear strength for reproducibility.

major comments (2)
  1. [§4] §4 (Experiments): the partition of degradation causes by multi-class vs. single-class-per-subject structure is not matched on label semantics (physiological vs. pathological signals) or dataset scale/stationarity; the observed split could therefore be produced by these unmatched covariates rather than the intended isolation of variability vs. shortcut mechanisms. A direct test (e.g., re-running the single-class experiments on physiological labels or vice versa) is needed to support the claim.
  2. [§5] §5 (Results and Discussion): no statistical tests or error bars are reported for the performance differences that underpin the task-type attribution; without them the separation between variability-dominant and shortcut-dominant regimes cannot be assessed for robustness.
minor comments (2)
  1. [Abstract, §3] Abstract and §3: the term 'shortcut learning' is used without a precise operational definition or reference to how it is measured (e.g., via specific ablation or feature attribution); add a short paragraph clarifying the metric.
  2. [Figures] Figure captions: several figures lack axis labels or legend details that would allow a reader to verify the cross-subject vs. within-subject comparison without returning to the main text.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and will revise the paper accordingly where feasible. We agree that statistical rigor should be improved and will add the requested analyses.

read point-by-point responses

  1. Referee: §4 (Experiments): the partition of degradation causes by multi-class vs. single-class-per-subject structure is not matched on label semantics (physiological vs. pathological signals) or dataset scale/stationarity; the observed split could therefore be produced by these unmatched covariates rather than the intended isolation of variability vs. shortcut mechanisms. A direct test (e.g., re-running the single-class experiments on physiological labels or vice versa) is needed to support the claim.

    Authors: We acknowledge that label semantics (physiological vs. pathological) and dataset characteristics are correlated with our multi-class vs. single-class categorization, as disease detection tasks are inherently single-label per subject while motor imagery/ERP tasks are multi-label. This correlation is intrinsic to the EEG problems studied and cannot be decoupled without fundamentally altering the tasks (e.g., forcing multi-class disease labels or single-class physiological tasks), which would change the scientific question. Our experiments hold model, preprocessing, and protocol fixed across tasks, and the patterns are consistent across multiple datasets per category. We will expand the discussion and limitations section to explicitly address these covariates and their potential influence. However, performing the suggested direct test would require new datasets and task reformulations outside the scope of the current work. revision: partial

  2. Referee: §5 (Results and Discussion): no statistical tests or error bars are reported for the performance differences that underpin the task-type attribution; without them the separation between variability-dominant and shortcut-dominant regimes cannot be assessed for robustness.

    Authors: We agree that reporting error bars and statistical tests is necessary to assess robustness. In the revised manuscript we will add standard deviation error bars (computed across random seeds or cross-validation folds) to all performance figures and tables. We will also include paired statistical tests (e.g., Wilcoxon signed-rank tests) comparing within-subject vs. cross-subject performance within each task category, with p-values reported to evaluate the significance of the observed differences. revision: yes

standing simulated objections not resolved
  • The request for a direct test by re-running single-class experiments on physiological labels (or vice versa), as this requires new data collection and task redesign not feasible within the current study.

Circularity Check

0 steps flagged

Empirical experiments with no circular derivation chain

full rationale

The paper reports results from a series of controlled experiments across EEG tasks to attribute cross-subject performance drops to inter-subject variability (in multi-class-per-subject tasks) versus shortcut learning (in single-class-per-subject tasks). No equations, fitted parameters renamed as predictions, self-citations as load-bearing premises, or ansatzes imported via prior work appear in the derivation. The central claims rest on experimental isolation and observation, which are externally falsifiable via replication on the released code and datasets. This is self-contained empirical work; no step reduces to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an empirical study relying on standard machine-learning evaluation practices for EEG; no free parameters, axioms, or invented entities are introduced in the abstract.

pith-pipeline@v0.9.0 · 5698 in / 1059 out tokens · 17098 ms · 2026-05-23T19:56:01.209081+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

51 extracted references · 51 canonical work pages · 2 internal anchors

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

    work page

  2. [2]

    Object classification from randomized eeg trials, 2020

    Hamad Ahmed, Ronnie B Wilbur, Hari M Bharadwaj, and Jeffrey Mark Siskind. Object classification from randomized eeg trials, 2020. URL https://arxiv.org/abs/2004.06046

    work page arXiv 2020

  3. [3]

    Deep learning techniques for classification of electroencephalogram (eeg) motor imagery (mi) signals: A review

    Hamdi Altaheri, Ghulam Muhammad, Mansour Alsulaiman, Syed Umar Amin, Ghadir Ali Altuwaijri, Wadood Abdul, Mohamed A Bencherif, and Mohammed Faisal. Deep learning techniques for classification of electroencephalogram (eeg) motor imagery (mi) signals: A review. Neural Computing and Applications, 35 0 (20): 0 14681--14722, 2023

    work page 2023

  4. [4]

    Ef-net: Mental state recognition by analyzing multimodal eeg-fnirs via cnn

    Aniqa Arif, Yihe Wang, Rui Yin, Xiang Zhang, and Ahmed Helmy. Ef-net: Mental state recognition by analyzing multimodal eeg-fnirs via cnn. Sensors, 24 0 (6): 0 1889, 2024

    work page 2024

  5. [5]

    A review on evaluating mental stress by deep learning using eeg signals

    Yara Badr, Usman Tariq, Fares Al-Shargie, Fabio Babiloni, Fadwa Al Mughairbi, and Hasan Al-Nashash. A review on evaluating mental stress by deep learning using eeg signals. Neural Computing and Applications, pp.\ 1--26, 2024

    work page 2024

  6. [6]

    An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling

    Shaojie Bai, J Zico Kolter, and Vladlen Koltun. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. arXiv preprint arXiv:1803.01271, 2018

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  7. [7]

    The berlin brain--computer interface: non-medical uses of bci technology

    Benjamin Blankertz, Michael Tangermann, Carmen Vidaurre, Siamac Fazli, Claudia Sannelli, Stefan Haufe, Cecilia Maeder, Lenny Ramsey, Irene Sturm, Gabriel Curio, et al. The berlin brain--computer interface: non-medical uses of bci technology. Frontiers in neuroscience, 4: 0 198, 2010

    work page 2010

  8. [8]

    Eeg classification of motor imagery using a novel deep learning framework

    Mengxi Dai, Dezhi Zheng, Rui Na, Shuai Wang, and Shuailei Zhang. Eeg classification of motor imagery using a novel deep learning framework. Sensors, 19 0 (3): 0 551, 2019

    work page 2019

  9. [9]

    Deep learning in fnirs: a review

    Condell Eastmond, Aseem Subedi, Suvranu De, and Xavier Intes. Deep learning in fnirs: a review. Neurophotonics, 9 0 (4): 0 041411--041411, 2022

    work page 2022

  10. [10]

    An attention-based deep learning approach for sleep stage classification with single-channel eeg

    Emadeldeen Eldele, Zhenghua Chen, Chengyu Liu, Min Wu, Chee-Keong Kwoh, Xiaoli Li, and Cuntai Guan. An attention-based deep learning approach for sleep stage classification with single-channel eeg. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 29: 0 809--818, 2021

    work page 2021

  11. [11]

    The application of deep learning algorithms for ppg signal processing and classification

    Filipa Esgalhado, Beatriz Fernandes, Valentina Vassilenko, Arnaldo Batista, and Sara Russo. The application of deep learning algorithms for ppg signal processing and classification. Computers, 10 0 (12): 0 158, 2021

    work page 2021

  12. [12]

    T-rep: Representation learning for time series using time-embeddings

    Archibald Fraikin, Adrien Bennetot, and St \'e phanie Allassonni \`e re. T-rep: Representation learning for time series using time-embeddings. ICLR, 2023

    work page 2023

  13. [13]

    Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals

    Ary L Goldberger, Luis AN Amaral, Leon Glass, Jeffrey M Hausdorff, Plamen Ch Ivanov, Roger G Mark, Joseph E Mietus, George B Moody, Chung-Kang Peng, and H Eugene Stanley. Physiobank, physiotoolkit, and physionet: components of a new research resource for complex physiologic signals. circulation, 101 0 (23): 0 e215--e220, 2000

    work page 2000

  14. [14]

    A convolutional neural network approach for classification of dementia stages based on 2d-spectral representation of eeg recordings

    Cosimo Ieracitano, Nadia Mammone, Alessia Bramanti, Amir Hussain, and Francesco C Morabito. A convolutional neural network approach for classification of dementia stages based on 2d-spectral representation of eeg recordings. Neurocomputing, 323: 0 96--107, 2019

    work page 2019

  15. [15]

    Emotion recognition in eeg signals using deep learning methods: A review

    Mahboobeh Jafari, Afshin Shoeibi, Marjane Khodatars, Sara Bagherzadeh, Ahmad Shalbaf, David L \'o pez Garc \' a, Juan M Gorriz, and U Rajendra Acharya. Emotion recognition in eeg signals using deep learning methods: A review. Computers in Biology and Medicine, pp.\ 107450, 2023

    work page 2023

  16. [16]

    Clocs: Contrastive learning of cardiac signals across space, time, and patients

    Dani Kiyasseh, Tingting Zhu, and David A Clifton. Clocs: Contrastive learning of cardiac signals across space, time, and patients. In International Conference on Machine Learning, pp.\ 5606--5615. PMLR, 2021

    work page 2021

  17. [17]

    Invasive brain computer interface for motor restoration in spinal cord injury: a systematic review

    Jordan J Levett, Lior M Elkaim, Farbod Niazi, Michael H Weber, Christian Iorio-Morin, Marco Bonizzato, and Alexander G Weil. Invasive brain computer interface for motor restoration in spinal cord injury: a systematic review. Neuromodulation: Technology at the Neural Interface, 2023

    work page 2023

  18. [18]

    A novel ensemble learning method using multiple objective particle swarm optimization for subject-independent eeg-based emotion recognition

    Rui Li, Chao Ren, Xiaowei Zhang, and Bin Hu. A novel ensemble learning method using multiple objective particle swarm optimization for subject-independent eeg-based emotion recognition. Computers in biology and medicine, 140: 0 105080, 2022

    work page 2022

  19. [19]

    Deep learning in ecg diagnosis: A review

    Xinwen Liu, Huan Wang, Zongjin Li, and Lang Qin. Deep learning in ecg diagnosis: A review. Knowledge-Based Systems, 227: 0 107187, 2021

    work page 2021

  20. [20]

    itransformer: Inverted transformers are effective for time series forecasting

    Yong Liu, Tengge Hu, Haoran Zhang, Haixu Wu, Shiyu Wang, Lintao Ma, and Mingsheng Long. itransformer: Inverted transformers are effective for time series forecasting. International Conference on Learning Representations, 2024

    work page 2024

  21. [21]

    Dice-net: a novel convolution-transformer architecture for alzheimer detection in eeg signals

    Andreas Miltiadous, Emmanouil Gionanidis, Katerina D Tzimourta, Nikolaos Giannakeas, and Alexandros T Tzallas. Dice-net: a novel convolution-transformer architecture for alzheimer detection in eeg signals. IEEE Access, 2023 a

    work page 2023

  22. [22]

    A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg

    Andreas Miltiadous, Katerina D Tzimourta, Theodora Afrantou, Panagiotis Ioannidis, Nikolaos Grigoriadis, Dimitrios G Tsalikakis, Pantelis Angelidis, Markos G Tsipouras, Euripidis Glavas, Nikolaos Giannakeas, et al. A dataset of scalp eeg recordings of alzheimer’s disease, frontotemporal dementia and healthy subjects from routine eeg. Data, 8 0 (6): 0 95, 2023 b

    work page 2023

  23. [23]

    Sleepeegnet: Automated sleep stage scoring with sequence to sequence deep learning approach

    Sajad Mousavi, Fatemeh Afghah, and U Rajendra Acharya. Sleepeegnet: Automated sleep stage scoring with sequence to sequence deep learning approach. PloS one, 14 0 (5): 0 e0216456, 2019

    work page 2019

  24. [24]

    An integrated brain-machine interface platform with thousands of channels

    Elon Musk et al. An integrated brain-machine interface platform with thousands of channels. Journal of medical Internet research, 21 0 (10): 0 e16194, 2019

    work page 2019

  25. [25]

    A comparative study of subject-dependent and subject-independent strategies for eeg-based emotion recognition using lstm network

    Debarshi Nath, Anubhav, Mrigank Singh, Divyashikha Sethia, Diksha Kalra, and S Indu. A comparative study of subject-dependent and subject-independent strategies for eeg-based emotion recognition using lstm network. In Proceedings of the 2020 4th International Conference on Compute and Data Analysis, pp.\ 142--147, 2020

    work page 2020

  26. [26]

    Subject independent emotion recognition from eeg using vmd and deep learning

    Pallavi Pandey and KR Seeja. Subject independent emotion recognition from eeg using vmd and deep learning. Journal of King Saud University-Computer and Information Sciences, 34 0 (5): 0 1730--1738, 2022

    work page 2022

  27. [27]

    Physionet: components of a new research resource for complex physiologic signals

    PhysioToolkit PhysioBank. Physionet: components of a new research resource for complex physiologic signals. Circulation, 101 0 (23): 0 e215--e220, 2000

    work page 2000

  28. [28]

    Multilayer perceptron and neural networks

    Marius-Constantin Popescu, Valentina E Balas, Liliana Perescu-Popescu, and Nikos Mastorakis. Multilayer perceptron and neural networks. WSEAS Transactions on Circuits and Systems, 8 0 (7): 0 579--588, 2009

    work page 2009

  29. [29]

    A deep learning-based comparative study to track mental depression from eeg data

    Avik Sarkar, Ankita Singh, and Rakhi Chakraborty. A deep learning-based comparative study to track mental depression from eeg data. Neuroscience Informatics, 2 0 (4): 0 100039, 2022

    work page 2022

  30. [30]

    Bci2000: a general-purpose brain-computer interface (bci) system

    Gerwin Schalk, Dennis J McFarland, Thilo Hinterberger, Niels Birbaumer, and Jonathan R Wolpaw. Bci2000: a general-purpose brain-computer interface (bci) system. IEEE Transactions on biomedical engineering, 51 0 (6): 0 1034--1043, 2004

    work page 2004

  31. [31]

    u bler, and Gernot R M \

    Reinhold Scherer, Josef Faller, Elisabeth VC Friedrich, Eloy Opisso, Ursula Costa, Andrea K \"u bler, and Gernot R M \"u ller-Putz. Individually adapted imagery improves brain-computer interface performance in end-users with disability. PloS one, 10 0 (5): 0 e0123727, 2015

    work page 2015

  32. [32]

    Deprnet: A deep convolution neural network framework for detecting depression using eeg

    Ayan Seal, Rishabh Bajpai, Jagriti Agnihotri, Anis Yazidi, Enrique Herrera-Viedma, and Ondrej Krejcar. Deprnet: A deep convolution neural network framework for detecting depression using eeg. IEEE Transactions on Instrumentation and Measurement, 70: 0 1--13, 2021

    work page 2021

  33. [33]

    u hmann, Benjamin Blankertz, Do-Won Kim, Jichai Jeong, Han-Jeong Hwang, and Klaus-Robert M \

    Jaeyoung Shin, Alexander von L \"u hmann, Benjamin Blankertz, Do-Won Kim, Jichai Jeong, Han-Jeong Hwang, and Klaus-Robert M \"u ller. Open access dataset for eeg+ nirs single-trial classification. IEEE Transactions on Neural Systems and Rehabilitation Engineering, 25 0 (10): 0 1735--1745, 2016

    work page 2016

  34. [34]

    u hmann, Do-Won Kim, Jan Mehnert, Han-Jeong Hwang, and Klaus-Robert M \

    Jaeyoung Shin, Alexander Von L \"u hmann, Do-Won Kim, Jan Mehnert, Han-Jeong Hwang, and Klaus-Robert M \"u ller. Simultaneous acquisition of eeg and nirs during cognitive tasks for an open access dataset. Scientific data, 5 0 (1): 0 1--16, 2018

    work page 2018

  35. [35]

    The two decades brainclinics research archive for insights in neurophysiology (tdbrain) database

    Hanneke van Dijk, Guido van Wingen, Damiaan Denys, Sebastian Olbrich, Rosalinde van Ruth, and Martijn Arns. The two decades brainclinics research archive for insights in neurophysiology (tdbrain) database. Scientific data, 9 0 (1): 0 333, 2022

    work page 2022

  36. [36]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, ukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in neural information processing systems, 30, 2017

    work page 2017

  37. [37]

    Ptb-xl, a large publicly available electrocardiography dataset

    Patrick Wagner, Nils Strodthoff, Ralf-Dieter Bousseljot, Dieter Kreiseler, Fatima I Lunze, Wojciech Samek, and Tobias Schaeffter. Ptb-xl, a large publicly available electrocardiography dataset. Scientific data, 7 0 (1): 0 1--15, 2020

    work page 2020

  38. [38]

    Timemixer: Decomposable multiscale mixing for time series forecasting

    Shiyu Wang, Haixu Wu, Xiaoming Shi, Tengge Hu, Huakun Luo, Lintao Ma, James Y Zhang, and Jun Zhou. Timemixer: Decomposable multiscale mixing for time series forecasting. ICLR, 2024 a

    work page 2024

  39. [39]

    Contrast everything: A hierarchical contrastive framework for medical time-series

    Yihe Wang, Yu Han, Haishuai Wang, and Xiang Zhang. Contrast everything: A hierarchical contrastive framework for medical time-series. Advances in Neural Information Processing Systems, 36, 2024 b

    work page 2024

  40. [40]

    Medformer: A multi-granularity patching transformer for medical time-series classification

    Yihe Wang, Nan Huang, Taida Li, Yujun Yan, and Xiang Zhang. Medformer: A multi-granularity patching transformer for medical time-series classification. arXiv preprint arXiv:2405.19363, 2024 c

    work page arXiv 2024

  41. [41]

    Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting

    Haixu Wu, Jiehui Xu, Jianmin Wang, and Mingsheng Long. Autoformer: Decomposition transformers with auto-correlation for long-term series forecasting. Advances in neural information processing systems, 34: 0 22419--22430, 2021

    work page 2021

  42. [42]

    TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis

    Haixu Wu, Tengge Hu, Yong Liu, Hang Zhou, Jianmin Wang, and Mingsheng Long. Timesnet: Temporal 2d-variation modeling for general time series analysis. arXiv preprint arXiv:2210.02186, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  43. [43]

    Manydg: Many-domain generalization for healthcare applications

    Chaoqi Yang, M Brandon Westover, and Jimeng Sun. Manydg: Many-domain generalization for healthcare applications. In The Eleventh International Conference on Learning Representations, 2022

    work page 2022

  44. [44]

    Subject-independent eeg classification based on a hybrid neural network

    Hao Zhang, Hongfei Ji, Jian Yu, Jie Li, Lingjing Jin, Lingyu Liu, Zhongfei Bai, and Chen Ye. Subject-independent eeg classification based on a hybrid neural network. Frontiers in Neuroscience, 17: 0 1124089, 2023

    work page 2023

  45. [45]

    Mindid: Person identification from brain waves through attention-based recurrent neural network

    Xiang Zhang, Lina Yao, Salil S Kanhere, Yunhao Liu, Tao Gu, and Kaixuan Chen. Mindid: Person identification from brain waves through attention-based recurrent neural network. Proceedings of the ACM on Interactive, Mobile, Wearable and Ubiquitous Technologies, 2 0 (3): 0 1--23, 2018

    work page 2018

  46. [46]

    Adversarial representation learning for robust patient-independent epileptic seizure detection

    Xiang Zhang, Lina Yao, Manqing Dong, Zhe Liu, Yu Zhang, and Yong Li. Adversarial representation learning for robust patient-independent epileptic seizure detection. IEEE journal of biomedical and health informatics, 24 0 (10): 0 2852--2859, 2020

    work page 2020

  47. [47]

    Informer: Beyond efficient transformer for long sequence time-series forecasting

    Haoyi Zhou, Shanghang Zhang, Jieqi Peng, Shuai Zhang, Jianxin Li, Hui Xiong, and Wancai Zhang. Informer: Beyond efficient transformer for long sequence time-series forecasting. In Proceedings of the AAAI conference on artificial intelligence, volume 35, pp.\ 11106--11115, 2021

    work page 2021

  48. [48]

    Semi-supervised learning for multi-label cardiovascular diseases prediction: a multi-dataset study

    Rushuang Zhou, Lei Lu, Zijun Liu, Ting Xiang, Zhen Liang, David A Clifton, Yining Dong, and Yuan-Ting Zhang. Semi-supervised learning for multi-label cardiovascular diseases prediction: a multi-dataset study. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    work page 2023

  49. [49]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should not add it explicitly Type <Return> for now, but then later remove the command n...

    work page

  50. [50]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@first@sw \@firstoftwo \@ifundefined NAT@b*@#2 \@firstoftwo @num @NAT@ctr \@secondoft...

    work page

  51. [51]

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibsetup #1 @NAT@ctr @ @openbib .11em \@plus.33em \@minus.07em 4000 4000 `\.\@m @bibit...

    work page arXiv