Hi, I'm Emadeldeen Eldele.

Currently, I am an Assistant Professor in the Department of Computer Science at Khalifa University, Abu Dhabi, United Arab Emirates.

Previously, I was a Research Scientist at A*STAR, Singapore, jointly appointed at the Institute for Infocomm Research (I²R) and the Centre for Frontier AI Research (CFAR), working on deep learning solutions for real-world time-series applications.

Education. I received my Ph.D. in Computer Science and Engineering from Nanyang Technological University (NTU), Singapore, and my B.Sc. and M.Sc. degrees in Computer Engineering from Tanta University, Egypt.

Latest Updates

DEC 2025One paper is accepted in the IEEE TPAMI.
NOV 2025One paper is accepted in AAAI 2026.
SEP 2025One paper is accepted in NeurIPS 2025.
SEP 2025One paper is accepted in the IEEE TKDE.
May 2025Our paper " Learning Soft Sparse Shapes for Efficient Time-Series Classification " has been accepted in ICML 2025.
May 2025Our paper " Counterfactual Contrastive Learning with Normalizing Flows for Robust Treatment Effect Estimation " has been accepted in ICML 2025.
Dec 2024Our paper " Hierarchical Classification Auxiliary Network for Time Series Forecasting" has been accepted in AAAI 2025.
Jul 2024Our survey paper " Label-efficient Time Series Representation Learning: A Review" has been accepted in the IEEE TAI.
May 2024Our paper "TSLANet: Rethinking Transformers for Time Series Representation Learning" has been accepted in ICML 2024.
April 2024Our paper "Bi-Hemisphere Interaction Convolutional Neural Network for Motor Imagery Classification" has been accepted in EMBC 2024.
Nov 2023Our paper "ECGTransForm: Empowering adaptive ECG arrhythmia classification framework with bidirectional transformer" has been accepted for publication in the Biomedical Signal Processing and Control.
Sep 2023I've received the prestigious 3rd Prize in the IEEE Engineering in Medicine and Biology Prize Paper Award (2023) for the AttnSleep paper.
Sep 2023Call for Papers: I'm serving as a Guest Editor in the Special Issue, entitled "Artificial Intelligence for Medical Sensing", in Sensors Journal.
Aug 2023Our paper "Self-supervised Contrastive Representation Learning for Semi-supervised Time-Series Classification" has been accepted for publication in the IEEE TPAMI.
Jul 2023Our paper " Contrastive Domain Adaptation for Time-Series via Temporal Mixup" has been accepted in the IEEE TAI.
May 2023Our paper "Source-Free Domain Adaptation with Temporal Imputation for Time Series Data" has been accepted in ACM KDD'23.
FEB 2023Our paper "ADATIME: A Benchmarking Suite for Domain Adaptation on Time Series Data" has been accepted in the ACM TKDD.
JAN 2023Our evaluation paper "Self-supervised Learning for Label-Efficient Sleep Stage Classification: A Comprehensive Evaluation" has been accepted in the IEEE TNSRE.

Research

Generalization, robustness, and reasoning in time-series AI systems.

Core Vision

Our research aims to develop AI systems for time-series data that are robust, generalizable, and capable of reasoning under uncertainty. Rather than optimizing for isolated benchmarks, we focus on models that remain stable across domains, adapt to distribution shifts, and support decision-making in real-world environments.

Key Principles

Generalization under Distribution Shift: Designing models that transfer across domains, sensors, and operating conditions.
Robust Representation Learning: Learning temporal abstractions that capture structure while remaining adaptable.
Uncertainty and Reliability: Explicit modeling of uncertainty to enable safe deployment in high-stakes settings.
Reasoning over Temporal Data: Moving beyond prediction toward structured abstraction and decision-oriented reasoning.

Long-Term Direction

Building on robust representation learning, our long-term objective is to develop reasoning-centric and adaptive systems that integrate temporal perception, structured reasoning (including LLM-based components), and closed-loop interaction. The overarching goal is to enable trustworthy AI systems that operate reliably over time.

Research Themes

Generalization & Domain Adaptation

Learning invariant mechanisms and transferable representations that remain stable across heterogeneous time-series domains.

Temporal Representation Learning

Structured modeling of temporal dynamics, multi-scale patterns, and adaptive segmentation for classification and forecasting tasks.

Reasoning over Time-Series

Bridging temporal abstractions with higher-level reasoning systems, including structured interfaces with large language models.

Trustworthy & Adaptive Systems

Uncertainty-aware learning, continual adaptation, and safe deployment in healthcare and industrial environments.

Selected Projects

Ongoing

Robust Time-Series Modeling under Distribution Shift

Developing invariant and domain-aware learning methods that enable stable transfer across datasets, devices, and operating conditions.

Domain Adaptation Generalization Causal Learning

Ongoing

Temporal Abstractions for Foundation & Transferable Models

Designing architectures that capture multi-scale patterns, distribution heterogeneity, and structural temporal dynamics for classification and forecasting.

Representation Learning Time Series Robust AI

Planned

Reasoning-Centric & Agentic Time-Series Systems

Integrating temporal perception with structured reasoning modules and adaptive decision-making pipelines for closed-loop systems in healthcare and industrial domains.

LLMs Agentic AI Trustworthy AI

Publications

2026

Time Series Domain Adaptation via Latent Invariant Causal Mechanism

Ruichu Cai; Junxian Huang; Zhenhui Yang; Zijian Li; Emadeldeen Eldele; Min Wu; Fuchun Sun

IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI)

PDF

Latent Causality Alignment (LCA) model, designed for time series domain adaptation, aims to transfer knowledge from a labeled source domain to an unlabeled target domain, particularly in complex, high-dimensional data such as videos where traditional causal modeling faces challenges.

@article{cai2026time,
      title   = {Time Series Domain Adaptation via Latent Invariant Causal Mechanism},
      author  = {Ruichu Cai, Junxian Huang, Zhenhui Yang, Zijian Li, Emadeldeen Eldele, Min Wu, Fuchun Sun},
      journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
      year    = {2026}
}

2026

EntroPE: Entropy-Guided Dynamic Patch Encoder for Time Series Forecasting

Sachith Abeywickrama, Emadeldeen Eldele, Min Wu, Xiaoli Li, Chau Yuen

Under Review

PDF

EntroPE is an entropy-guided dynamic patching framework for time series transformers that detects natural temporal transitions using conditional entropy, preserves short-term dependencies, and achieves better accuracy and efficiency than fixed patching across forecasting, classification, and anomaly detection tasks.

@article{entrope,
      title   = {EntroPE: Entropy-Guided Dynamic Patch Encoder for Time Series Forecasting},
      author  = {Sachith Abeywickrama, Emadeldeen Eldele, Min Wu, Xiaoli Li, Chau Yuen},
      journal = {arXiv preprint arXiv:2509.26157}, 
      year    = {2025}
}

2026

A Unified Shape-Aware Foundation Model for Time Series Classification

Zhen Liu; Yucheng Wang; Boyuan Li; Junhao Zheng; Emadeldeen Eldele;, Min Wu; Qianli Ma;

The Fortieth AAAI Conference on Artificial Intelligence [Acceptance Rate: 23.4%]

PDF

@inproceedings{UniShape,
      title     = {A Unified Shape-Aware Foundation Model for Time Series Classification},
      author    = {Zhen Liu, Yucheng Wang, Boyuan Li, Junhao Zheng, Emadeldeen Eldele, Min Wu, Qianli Ma},
      booktitle = {The Fortieth AAAI Conference on Artificial Intelligence},
      year      = {2026}
}

2026

Evidentially Calibrated Source-Free Time-Series Domain Adaptation With Temporal Imputation

Mohamed Ragab; Peiliang Gong; Emadeldeen Eldele; Wenyu Zhang; Min Wu; and Chuan-Sheng Foo

IEEE Transactions on Knowledge and Data Engineering (TKDE)

Link

MAsk And imPUte (MAPU) and Evidential-MAPU (E-MAPU) are novel methods for Source-Free Domain Adaptation (SFDA) in time series analysis. SFDA enables a model trained on a labeled source dataset to adapt to an unlabeled target dataset without source data access for privacy preservation.

@article{ragab2026evidentially,
      title   = {Evidentially Calibrated Source-Free Time-Series Domain Adaptation With Temporal Imputation},
      author  = {Ragab, Mohamed and Gong, Peiliang and Eldele, Emadeldeen and Zhang, Wenyu and Wu, Min and Foo, Chuan-Sheng and Zhang, Daoqiang and Li, Xiaoli and Chen, Zhenghua},
      journal = {IEEE Transactions on Knowledge and Data Engineering}, 
      year    = {2026}
}

2025

Adapting llms to time series forecasting via temporal heterogeneity modeling and semantic alignment

Yanru Sun, Emadeldeen Eldele, Zongxia Xie, Yucheng Wang, Wenzhe Niu, Qinghua Hu, Chee Keong Kwoh, Min Wu

Under Review

PDF

TALON is a unified LLM-based time series forecasting framework that addresses temporal heterogeneity and the modality gap by using a heterogeneous temporal encoder and semantic alignment module, achieving up to 11% MSE improvement over state-of-the-art methods across seven real-world benchmarks.

@article{talon,
      title   = {Adapting LLMs to Time Series Forecasting via Temporal Heterogeneity Modeling and Semantic Alignment},
      author  = {Yanru Sun, Emadeldeen Eldele, Zongxia Xie, Yucheng Wang, Wenzhe Niu, Qinghua Hu, Chee Keong Kwoh, Min Wu},
      journal = {arXiv preprint arXiv:2508.07195}, 
      year    = {2025}
}

2025

Unifault: A fault diagnosis foundation model from bearing data

Emadeldeen Eldele, Mohamed Ragab, Xu Qing, Zhenghua Chen, Min Wu, Xiaoli Li, Jay Lee

Under Review

PDF

UniFault is a foundation model for machine fault diagnosis that overcomes dataset heterogeneity through data harmonization, univariate unification, and cross-domain temporal fusion, achieving state-of-the-art and strong few-shot performance after pretraining on 6.9 million samples from diverse FD datasets.

@article{unifault,
      title   = {Unifault: A fault diagnosis foundation model from bearing data},
      author  = {Emadeldeen Eldele, Mohamed Ragab, Xu Qing, Zhenghua Chen, Min Wu, Xiaoli Li, Jay Lee},
      journal = {arXiv preprint arXiv:2504.01373}, 
      year    = {2025}
}

2025

Learning Pattern-Specific Experts for Time Series Forecasting Under Patch-level Distribution Shift

Yanru Sun; Zongxia Xie; Emadeldeen Eldele; Dongyue Chen; Qinghua Hu; and Min Wu

Conference on Neural Information Processing Systems(NeurIPS) [Acceptance Rate: 24.52%]

PDF

TFPS introduces a novel deep learning framework that tackles non-stationary time series by using a Mixture of Experts (MoE) approach, dynamically routing data "patches" (segments) to specialized experts that model unique temporal patterns, thereby overcoming limitations of single models in handling shifting data distributions and achieving state-of-the-art accuracy on complex forecasting tasks

@inproceedings{TFPS,
      title     = {Learning Pattern-Specific Experts for Time Series Forecasting Under Patch-level Distribution Shift},
      author    = {Yanru Sun, Zongxia Xie, Emadeldeen Eldele, Dongyue Chen, Qinghua Hu, Min Wu},
      booktitle = {The Thirty-ninth Annual Conference on Neural Information Processing Systems},
      year      = {2025}
}

2025

From maxwell’s equations to artificial intelligence: The evolution of physics-guided AI in nanophotonics and electromagnetics

Omar A.M. Abdelraouf; Abdulrahman M.A. Ahmed; Emadeldeen Eldele; and Ahmed A. Omar

Optics & Laser Technology [Impact Factor (2025): 5.0]

Link

This review paper highlights how AI-augmented, physics-guided frameworks are revolutionizing electromagnetic and nanophotonic design by enabling efficient optimization, accelerating simulations, and surpassing traditional trial-and-error methods, while outlining key advances, challenges, and future opportunities for next-generation photonic technologies.

@article{ABDELRAOUF2025113828,
      title     = {From maxwell’s equations to artificial intelligence: The evolution of physics-guided AI in nanophotonics and electromagnetics},
      author    = {Omar A.M. Abdelraouf and Abdulrahman M.A. Ahmed and Emadeldeen Eldele and Ahmed A. Omar},
      journal   = {Optics & Laser Technology},
      volume    = {192},
      pages     = {113828},
      year      = {2025},
      doi       = {https://doi.org/10.1016/j.optlastec.2025.113828}
}

2025

Learning Soft Sparse Shapes for Efficient Time-Series Classification

Zhen Liu; Yicheng Luo; Boyuan Li; Emadeldeen Eldele; Min Wu; and Qianli Ma

International Conference on Machine Learning (ICML) [Acceptance Rate: 26.9%]

PDF

SoftShape improves time series classification with soft shape sparsification and learning, using a learnable router for intra-shape patterns and a shared network for inter-shape patterns, outperforming existing methods with interpretable results.

@inproceedings{softshape,
      title     = {Learning Soft Sparse Shapes for Efficient Time-Series Classification},
      author    = {Zhen Liu, Yicheng Luo, Boyuan Li, Emadeldeen Eldele, Min Wu, Qianli Ma},
      booktitle = {International Conference on Machine Learning},
      year      = {2025}
}

2025

Counterfactual Contrastive Learning with Normalizing Flows for Robust Treatment Effect Estimation

Jiaxuan Zhang; Emadeldeen Eldele; Fuyuan Cao; Yang Wang; Xiaoli Li; and Jiye Liang

International Conference on Machine Learning (ICML) [Acceptance Rate: 26.9%]

PDF

FCCL estimates Individual Treatment Effects by generating diffeomorphic counterfactuals and aligning samples to reduce distribution shift, guided by a theoretical bound linking estimation error to factual prediction and representation distances.

@inproceedings{fccl,
      title     = {Counterfactual Contrastive Learning with Normalizing Flows for Robust Treatment Effect Estimation},
      author    = {Jiaxuan Zhang, Emadeldeen Eldele, Fuyuan Cao, Yang Wang, Xiaoli Li, Jiye Liang},
      booktitle = {International Conference on Machine Learning},
      year      = {2025}
}

2025

Hierarchical Classification Auxiliary Network for Time Series Forecasting

Yanru Sun; Zongxia Xie; Dongyue Chen; Emadeldeen Eldele; and Qinghua Hu

AAAI Conference on Artificial Intelligence (AAAI) [Acceptance Rate: 23.4%]

PDF

This paper tackles the over-smoothing issue in time-series forecasting caused by Mean Squared Error (MSE) loss, presenting an innovative solution that enhances feature diversity by reframing the problem as a hierarchical classification task. The proposed Hierarchical Classification Auxiliary Network (HCAN) integrates Hierarchy-Aware Attention, an Uncertainty-Aware Classifier, and Hierarchical Consistency Loss to capture both fine and coarse-grained time-series representations.

@inproceedings{hcan,
      title     = {Hierarchical Classification Auxiliary Network for Time Series Forecasting},
      author    = {Yanru Sun, Zongxia Xie, Dongyue Chen, Emadeldeen Eldele, Qinghua Hu},
      booktitle = {AAAI Conference on Artificial Intelligence},
      year      = {2025}
}

2024

Bi-hemisphere Interaction Convolutional Neural Network for Motor Imagery Classification

Xiaohao Lin; Emadeldeen Eldele; Zhenghua Chen; Min Wu; Han Wei Ng; and Cuntai Guan

IEEE Engineering in Medicine and Biology Society (EMBC)

Link

The paper introduces BrainCoNet, a CNN architecture that models inter-hemispheric functional connectivity using Channel Average Referencing and cosine similarity for enhanced EEG-based motor imagery classification. It simulates bi-hemispheric brain interactions and establishes a biologically inspired framework for decoding motor imagery signals.

@inproceedings{bihemisphere,
      title     = {Bi-hemisphere Interaction Convolutional Neural Network for Motor Imagery Classification}, 
      author    = {Lin, Xiaohao and Eldele, Emadeldeen and Chen, Zhenghua and Wu, Min and Ng, Han Wei and Guan, Cuntai},
      booktitle = {2024 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)}, 
      year      = {2024},
      pages     = {1-4},
      doi       = {10.1109/EMBC53108.2024.10782755}
}

2024

Label-efficient Time Series Representation Learning: A Review

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; and Xiaoli Li

IEEE Transactions on Artificial Intelligence (TAI)

PDF

This survey paper presents a new taxonomy to categorize existing approaches that deal with the scarcity of labeled data in time series data such as transfer learning, self-supervised learning, and semi-supervised learning. The paper also reviews recent advances in each approach, discusses their limitations, and suggests future directions for further progress in the field.

@article{emadeldeen2024labelefficient,
      title   = {Label-efficient Time Series Representation Learning: A Review},
      author  = {Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Kwoh, Chee Keong and Li, Xiaoli},
      journal = {IEEE Transactions on Artificial Intelligence}, 
      year    = {2024}
}

2024

TSLANet: Rethinking Transformers for Time Series Representation Learning

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Xiaoli Li

International Conference on Machine Learning (ICML) [Acceptance Rate: 27.5%]

PDF

This paper presents TSLANet, a novel, efficient approach combining convolutional operations with adaptive spectral analysis. TSLANet posits a universal approach tailored for time series data, which is crucial for a myriad of applications.

@inproceedings{tslanet,
      title     = {TSLANet: Rethinking Transformers for Time Series Representation Learning},
      author    = {Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Li, Xiaoli},
      booktitle = {International Conference on Machine Learning},
      year      = {2024}
}

2024

ECGTransForm: Empowering adaptive ECG arrhythmia classification framework with bidirectional transformer

Hany El-Ghaish and Emadeldeen Eldele;

Biomedical Signal Processing and Control [IF: 5.1]

Link

This paper presents ECGTransForm, a deep learning framework for ECG arrhythmia classification, featuring a novel Bidirectional Transformer mechanism and Multi-scale Convolutions for effective spatial and temporal feature extraction. The framework includes a Context-Aware Loss to handle class imbalance in ECG data, demonstrating superior performance in arrhythmia diagnosis.

@ARTICLE{ecgTransForm,
      title = {ECGTransForm: Empowering adaptive ECG arrhythmia classification framework with bidirectional transformer},
      journal = {Biomedical Signal Processing and Control},
      volume = {89},
      pages = {105714},
      year = {2024},
      issn = {1746-8094},
      doi = {https://doi.org/10.1016/j.bspc.2023.105714}, 
      url = {https://www.sciencedirect.com/science/article/pii/S1746809423011473},
      author = {Hany El-Ghaish and Emadeldeen Eldele}
}

2023

Self-supervised Contrastive Representation Learning for Semi-supervised Time-Series Classification

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; Xiaoli Li; Cuntai Guan

IEEE Transactions on Pattern Analysis and Machine Intelligence(TPAMI) [IF: 24.314]

PDF

This work proposes a novel Time-Series semi-supervised representation learning framework, CA-TCC, that learns representations given few labeled data using contrastive learning. CA-TCC builds on top of our TS-TCC work by leveraging the robust pseudo labels from the fine-tuned TS-TCC model in a supervised contrastive loss.

@ARTICLE{catcc,
     author={Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli and Guan, Cuntai},
     journal={IEEE Transactions on Pattern Analysis and Machine Intelligence}, 
     title={Self-Supervised Contrastive Representation Learning for Semi-Supervised Time-Series Classification}, 
     year={2023},
     volume={45},
     number={12},
     pages={15604-15618},
     doi={10.1109/TPAMI.2023.3308189}
}

2023

Source-Free Domain Adaptation with Temporal Imputation for Time Series Data

Mohamed Ragab; Emadeldeen Eldele; Min Wu; Chuan-Sheng Foo; Xiaoli Li; and Zhenghua Chen

ACM Conference On Knowledge Discovery and Data Mining (SIGKDD) [Acceptance rate: 22.1%]

Link

The paper introduces a novel source-free domain adaptation method called MAsk and imPUte (MAPU) for time series data. MAPU can effectively capture the temporal dynamics present in time series data without requiring access to the source data during adaptation. The method involves temporal masking and a temporal imputation network to ensure sequence consistency between the source and target domains. By incorporating temporal adaptation capability into existing source-free methods, MAPU achieves significant performance improvements in real-world time series adaptation tasks.

@inproceedings{MAPU,
   author = {Ragab, Mohamed and Eldele, Emadeldeen and Wu, Min and Foo, Chuan-Sheng and Li, Xiaoli and Chen, Zhenghua},
   title = {Source-Free Domain Adaptation with Temporal Imputation for Time Series Data},
   year = {2023},
   isbn = {9798400701030},
   publisher = {Association for Computing Machinery},
   address = {New York, NY, USA},
   url = {https://doi.org/10.1145/3580305.3599507},
   doi = {10.1145/3580305.3599507},
   booktitle = {Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining},
   pages = {1989–1998},
   numpages = {10},
   keywords = {time series data, temporal imputation, source-free domain adaptation, privacy preserving},
   location = {Long Beach, CA, USA},
   series = {KDD '23}
}

2023

Contrastive Domain Adaptation for Time-Series via Temporal Mixup

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; and Xiaoli Li

IEEE Transactions on Artificial Intelligence (TAI)

PDF

This work proposes a novel lightweight contrastive domain adaptation framework called CoTMix for time-series data, which uses contrastive learning solely to mitigate the distribution shift across different domains. We propose a novel temporal mixup strategy to generate intermediate augmented views for source and target domains, which are then leveraged for contrastive learning, which gradually pushes both domains towards a common intermediate space, while learning about their temporal relations.

@ARTICLE{cotmix,
     author={Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli},
     journal={IEEE Transactions on Artificial Intelligence}, 
     title={Contrastive Domain Adaptation for Time-Series Via Temporal Mixup}, 
     year={2024},
     volume={5},
     number={3},
     pages={1185-1194},
     doi={10.1109/TAI.2023.3293473}
}

2023

ADATIME: A Benchmarking Suite for Domain Adaptation on Time Series Data

Mohamed Ragab*; Emadeldeen Eldele* ; Wee Ling Tan; Chuan-Sheng Foo; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; Xiaoli Li (* Equal Contribution)

ACM Transactions on Knowledge Discovery from Data (TKDD) [IF: 2.713]

Link

The paper introduces ADATIME, a benchmarking evaluation suite that systematically evaluates different domain adaptation methods on time series data, standardizing the backbone network architectures, evaluation schemes, and benchmarking datasets. The results suggest that visual domain adaptation methods are competitive with methods specifically developed for time series data, and realistic model selection approaches can be used for hyper-parameter selection.

@article{adatime,
      author = {Ragab, Mohamed and Eldele, Emadeldeen and Tan, Wee Ling and Foo, Chuan-Sheng and Chen, Zhenghua and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli},
      title = {ADATIME: A Benchmarking Suite for Domain Adaptation on Time Series Data},
      year = {2023},
      issue_date = {September 2023},
      publisher = {Association for Computing Machinery},
      address = {New York, NY, USA},
      volume = {17},
      number = {8},
      issn = {1556-4681},
      url = {https://doi.org/10.1145/3587937},
      doi = {10.1145/3587937},
      journal = {ACM Trans. Knowl. Discov. Data},
      month = {may},
      articleno = {106},
      numpages = {18}
}

2023

Self-supervised Learning for Label-Efficient Sleep Stage Classification: A Comprehensive Evaluation

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; and Xiaoli Li

IEEE Transactions on Neural Systems and Rehabilitation Engineering (TNSRE) [IF: 4.528]

Link

This paper discusses the challenge of limited labeled data for deep learning models used in EEG-based sleep stage classification (SSC) and explores the effectiveness of self-supervised learning (SSL) to improve the performance of SSC models in scenarios with few labeled data. We conduct a study on three SSC datasets and find that fine-tuning pretrained models with only 5% labeled data can achieve competitive performance and that SSL pretraining can improve the models' robustness to data imbalance and domain shift issues.

@article{emadeldeen2023eval,
      title   = {Self-supervised Learning for Label-Efficient Sleep Stage Classification: A Comprehensive Evaluation},
      author  = {Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Kwoh, Chee Keong and Li, Xiaoli},
      journal = {IEEE Transactions on Neural Systems and Rehabilitation Engineering}, 
      year    = {2023},
      volume  = {31},
      pages   = {1333-1342},
      doi     = {10.1109/TNSRE.2023.3245285}
}

2022

ADAST: Attentive Cross-domain EEG-based Sleep Staging Framework with Iterative Self-Training

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; Xiaoli Li; Cuntai Guan

IEEE Transactions on Emerging Topics in Computational Intelligence (TETCI) [IF:4.851]

PDF

This paper proposes a new adversarial learning framework called ADAST to address the cross-domain shift problem in automatic sleep staging. ADAST preserves domain-specific features via unshared attention model, and incorporates an iterative self-training strategy to improve the target domain class alignment.

@ARTICLE{adast,
      author  = {Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli and Guan, Cuntai},
      journal = {IEEE Transactions on Emerging Topics in Computational Intelligence}, 
      title   = {ADAST: Attentive Cross-Domain EEG-Based Sleep Staging Framework With Iterative Self-Training}, 
      year    = {2022},
      pages   = {1-12},
      doi     = {10.1109/TETCI.2022.3189695}
}

2022

Self-supervised Autoregressive Domain Adaptation for Time Series Data

Mohamed Ragab; Emadeldeen Eldele; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; Xiaoli Li

IEEE Transactions on Neural Networks and Learning Systems (TNNLS) [IF:14.255]

Link

This paper describes a new approach to address the domain shift problem in time series data using unsupervised domain adaptation called Self-supervised Autoregressive Domain Adaptation (SLARDA). In SLARDA, we propose a self-supervised learning module that improves the transferability of source features, in addition to a novel autoregressive domain adaptation technique to incorporate temporal dependency in adversarial training. Last, we incorporate an ensemble teacher model to align the class-wise distribution in the target domain using confident pseudo labels.

@article{slarda,
      author  = {Ragab, Mohamed and Eldele, Emadeldeen and Chen, Zhenghua and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli},
      journal = {IEEE Transactions on Neural Networks and Learning Systems}, 
      title   = {Self-supervised Autoregressive Domain Adaptation for Time Series Data}, 
      year    = {2022},
      pages   = {1-11},
      doi     = {10.1109/TNNLS.2022.3183252}
}

2022

Conditional Contrastive Domain Generalization for Fault Diagnosis

Mohamed Ragab; Zhenghua Chen; Wenyu Zhang; Emadeldeen Eldele; Min Wu; Chee-Keong Kwoh; Xiaoli Li

IEEE Transactions on Instrumentation and Measurement (TIM) [IF: 5.332]

Link

This paper proposes a new domain generalization approach called CCDG for fault diagnosis of rolling machinery that can learn environment-independent representation among data collected from different environments. This approach attempts to maximize the mutual information of similar classes across different source domains while minimizing mutual information among different classes to learn domain-independent class representation.

@ARTICLE{CCDG,
      author  = {Ragab, Mohamed and Chen, Zhenghua and Zhang, Wenyu and Eldele, Emadeldeen and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli},
      journal = {IEEE Transactions on Instrumentation and Measurement}, 
      title   = {Conditional Contrastive Domain Generalization for Fault Diagnosis}, 
      year    = {2022},
      volume  = {71},
      pages   = {1-12},
      doi     = {10.1109/TIM.2022.3154000}
}

2021

Time-Series Representation Learning via Temporal and Contextual Contrasting

Emadeldeen Eldele; Mohamed Ragab; Zhenghua Chen; Min Wu; Chee-Keong Kwoh; Xiaoli Li; Cuntai Guan

International Joint Conferences on Artificial Intelligence (IJCAI-21) [Acceptance Rate: 13.9%]

PDF

This paper proposes a self-supervised Time-Series representation learning framework via Temporal and Contextual Contrasting (TS-TCC), which learns representations from unlabeled data. First, we transform the raw data into two different yet correlated views using weak and strong augmentations. Next, a novel cross-view temporal contrasting module and a contextual contrasting module are proposed to learn robust temporal representations and further learn discriminative representations from the generated augmentated views.

@inproceedings{tstcc,
      title     = {Time-Series Representation Learning via Temporal and Contextual Contrasting},
      author    = {Eldele, Emadeldeen and Ragab, Mohamed and Chen, Zhenghua and Wu, Min and Kwoh, Chee Keong and Li, Xiaoli and Guan, Cuntai},
      booktitle = {Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence, {IJCAI-21}},
      pages     = {2352--2359},
      year      = {2021}
}

2021

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; Cuntai Guan

IEEE Transactions on Neural Systems and Rehabilitation Engineering (TNSRE) [IF: 4.528]

Link

This paper proposes AttnSleep; a deep learning architecture for automatic sleep stage classification using single channel EEG signals. AttnSleep consists of a feature extraction module based on a powerful feature extractor, followed by a temporal context encoder module that uses a multi-head attention mechanism to capture temporal dependencies among the features. It also addresses the class-imbalance problem via a class-aware loss.

@article{attnsleep,
      author  = {Eldele, Emadeldeen and Chen, Zhenghua and Liu, Chengyu and Wu, Min and Kwoh, Chee-Keong and Li, Xiaoli and Guan, Cuntai},
      journal = {IEEE Transactions on Neural Systems and Rehabilitation Engineering}, 
      title   = {An Attention-based Deep Learning Approach for Sleep Stage Classification with Single-Channel EEG}, 
      year    = {2021},
      volume  = {29},
      pages   = {809-818},
      doi     = {10.1109/TNSRE.2021.3076234}
}

2017

Enhanced framework for miRNA target prediction

Emadeldeen Eldele; Sherin M. El-Gokhy; Mohamed T. Faheem Saidahmed

12th International Conference on Computer Engineering and Systems (ICCES)

Link

This paper proposes an enhanced computational framework for predicting miRNA targets that uses structural, positional, and thermodynamic features extracted from miRNA-target alignment. The framework selects linearly separable features using eigenvalue analysis and applies them to a random forest classifier, achieving better results than existing tools.

@article{eldele_mirna,
      author    = {Eldele, Emadeldeen and El-Gokhy, Sherin M. and Saidahmed, Mohamed T. Faheem},
      booktitle = {2017 12th International Conference on Computer Engineering and Systems (ICCES)},
      title     = {Enhanced framework for miRNA target prediction}, 
      year      = {2017},
      pages     = {544-549},
      doi       = {10.1109/ICCES.2017.8275367}
}

Experience

Work Experience

Aug 2025 – Present

Assistant Professor

Khalifa University, UAE

Department of Computer Science.

Responsible for teaching undergraduate and graduate courses; supervising students; publishing in top-tier conferences and journals; and applying for grants.

May 2023 – Jul 2025

Research Scientist

CFAR & I²R, A*STAR, Singapore

Worked on representation learning and sequence modeling for time-series analysis and forecasting.

Developed unified frameworks for industrial fault diagnosis and remaining useful life (RUL) prediction.

Aug 2019 – May 2023

Research Scholar

Institute for Infocomm Research (I²R), A*STAR, Singapore

Developed label-efficient and self-supervised learning algorithms for time-series data.

Worked on domain adaptation and sleep stage classification models.

Education

Aug 2019 – Apr 2023

Ph.D. in Computer Science and Engineering

Nanyang Technological University (NTU), Singapore

Thesis: "Towards Robust and Label-efficient Time-Series Representation Learning".

Jan 2014 – Jun 2018

M.Sc. in Computer Science and Engineering

Tanta University, Egypt

Developed a machine learning framework for predicting miRNA targets in mRNA sequences.

Sep 2012 – May 2013

Cloud Architecture Diploma

Information Technology Institute (ITI), Egypt

Specialized in virtualization, storage management, and cloud system design.

May 2007 – Jul 2012

B.Sc. in Computer Science and Engineering

Tanta University, Egypt

Studied core topics in computer engineering, including neural networks and object-oriented programming.

Teaching

COSC 114: Introduction to Computing (Python)

Khalifa University (2025–Present)

Foundations of programming in Python with emphasis on problem solving and data-centric thinking.

COSC 201: Computer Systems Organization

Khalifa University (2025–Present)

Computer architecture, instruction sets, and memory hierarchy.

K6312: Information Mining & Analysis

NTU, Singapore – Guest Lecturer, 2024

Graduate module on data mining and analytical methods for large-scale data.

Teaching Assistant Roles

NTU & Tanta University (2013–2021)

Supported teaching in digital logic, object-oriented programming, data analytics, neural networks, and computer graphics.

Our Team

Meet the researchers behind the work.

Graduate Students

Asmaa Chehab

PhD Student

Working on representation learning for healthcare time-series data.

Get in Touch

I am always open to discussing new collaborations and research opportunities.

emadeldeenahmed.eldele@ku.ac.ae

Office

Room D04204, D Building
Main Campus
Khalifa University