I am currently a Postdoctoral fellow at Simula Metropolitan Center for Digital Engineering (SimulaMet) and part of the DRIVE and MISHMASH projects.
My background is on computer science, specializing in data mining and scientific computing. I am especially intrigued by unsupervised machine learning due to its capacity to discover patterns that are not predefined and relationships we did not even know to look for.
Education
2022 – 2026Ph.D. in Engineering Science, Simula Metropolitan Center for Digital Engineering
2020 – 2022M.Sc. in Computer Science, National and Kapodistrian University of Athens
2016 – 2020B.Sc. in Informatics & Telecommunications, National and Kapodistrian University of Athens
Christos Chatzis Simula Metropolitan Center for Digital Engineering, Oslo Metropolitan University, 2026
@phdthesis{01a133ccbb53-8f041381-64ce-41ce-95ef-1bb786b77a23,
author = {Christos Chatzis},
isbn = {9788283648355},
nva_api = {https://api.nva.unit.no/publication/01a133ccbb53-8f041381-64ce-41ce-95ef-1bb786b77a23},
pages = {269},
school = {Simula Metropolitan Center for Digital Engineering, Oslo Metropolitan University},
title = {Time-aware methods for detecting and analyzing patterns in time-evolving data},
url = {https://api.nva.unit.no/publication/01a133ccbb53-8f041381-64ce-41ce-95ef-1bb786b77a23},
year = {2026},
}
@article{erdos_extracting_2026,
doi = {10.1371/journal.pcbi.1014486},
author = {Erdős, Balázs AND Chatzis, Christos AND Thorsen, Jonathan AND Stokholm, Jakob AND Smilde, Age K. AND Rasmussen, Morten A. AND Acar, Evrim},
journal = {PLOS Computational Biology},
title = {Extracting host-specific developmental signatures from longitudinal microbiome data},
year = {2026},
volume = {22},
url = {https://doi.org/10.1371/journal.pcbi.1014486},
pages = {1-21},
number = {7},
code = {https://github.com/blzserdos/parafac2_microbiome},
}
@inproceedings{11226087,
author = {Chatzis, Christos and Schenker, Carla and Cohen, Jérémy E. and Acar, Evrim},
booktitle = {2025 33rd European Signal Processing Conference (EUSIPCO)},
title = {dCMF: Learning Interpretable Evolving Patterns from Temporal Multiway Data},
year = {2025},
volume = {},
number = {},
pages = {2122-2126},
doi = {10.23919/EUSIPCO63237.2025.11226087},
code = {https://github.com/cchatzis/temp-dcmf-repo},
}
@article{Chatzis2025,
author = {Christos Chatzis and Carla Schenker and Max Pfeffer and Evrim Acar},
title = {tPARAFAC2: tracking evolving patterns in (incomplete) temporal data},
journal = {Data Mining and Knowledge Discovery},
year = {2025},
volume = {39},
number = {6},
doi = {10.1007/s10618-025-01122-6},
code = {https://github.com/cchatzis/tPARAFAC2-latest},
}
Christos Chatzis, Max Pfeffer, Pedro Lind, Evrim Acar 2023 IEEE 33rd International Workshop on Machine Learning for Signal Processing (MLSP), 2023
@inproceedings{10285943,
author = {Chatzis, Christos and Pfeffer, Max and Lind, Pedro and Acar, Evrim},
booktitle = {2023 IEEE 33rd International Workshop on Machine Learning for Signal Processing (MLSP)},
title = {A Time-Aware Tensor Decomposition for Tracking Evolving Patterns},
year = {2023},
volume = {},
number = {},
pages = {1-6},
keywords = {Tensors;Frequency modulation;Conferences;Machine learning;Signal processing;Noise measurement;Matrix decomposition;Tensor decomposition;PARAFAC2;Time-evolving data},
doi = {10.1109/MLSP55844.2023.10285943},
}
Christos Chatzis National and Kapodistrian University of Athens, 2022
@masterthesis{uoadl:3197953,
bibtex_entry = {masterthesis},
year = {2022},
school = {National and Kapodistrian University of Athens},
author = {Christos Chatzis},
title = {A Least Squares Formulation for
Multilinear PCA and LDA},
url = {https://pergamos.lib.uoa.gr/uoa/dl/object/3197953},
}
Talks
Contributed talkRevealing Subject-Specific Temporal Patterns from Longitudinal Data
August 2026•1st KDD workshop on Interplay of Classical Tensor Methods and Foundation Models•Jeju, Korea
Temporal multivariate data is ubiquitous in many domains, collected over time at planned visits in longitudinal cohorts or at specific intervals in challenge tests. While the analysis of such data often focuses on revealing temporal patterns common across subjects, subject-specific differences in those patterns hold the promise to enhance our understanding of underlying mechanisms and facilitate personalized approaches. However, reliably extracting such temporal patterns from longitudinal multivariate data remains an open challenge. In this work, we demonstrate that coupled matrix factorizations (CMF) and PARAFAC2 are effective tools for capturing such subject-specific temporal profiles, focusing on two novel applications. In longitudinal metabolomics, CMF and PARAFAC2 reveal differences in postprandial metabolic responses of individuals according to body mass index (BMI) and insulin resistance (IR) measures. In sensitization data analysis, both methods capture differences in temporal trajectories of children according to delivery mode, insights that CANDECOMP/PARAFAC (CP), the current state of the art, fails to uncover. We also demonstrate the reproducibility and replicability of the extracted patterns. This is joint work with David Horner, Rasmus Bro, Ann-Marie M Schoos, Morten Arendt Rasmussen and Evrim Acar.
PosterRevealing Subject-Specific Temporal Patterns from Longitudinal Data
July 2026•Greeks in AI 2026•Athens, Greece
**Motivation:** Temporal multivariate data is ubiquitous in many domains, for instance, being collected over time at planned visits (every few months/years) in longitudinal cohorts, or every few minutes/hours in challenge tests. The analysis of such data often focuses on revealing the underlying temporal patterns common across subjects. However, there are subject-specific differences in temporal patterns, which hold the promise to enhance our understanding of underlying mechanisms and facilitate personalized approaches. Nevertheless, extracting subject-specific temporal patterns from longitudinal multivariate data reliably is an open challenge. **Results:** We introduce coupled matrix factorizations (CMF) as effective tools to capture subject-specific temporal patterns focusing on two novel applications: analysis of longitudinal metabolomics data and sensitization data. Our analysis shows that CMF models reliably capture subject-specific (shape) differences in temporal patterns revealing further in-sights compared to the state of the art. In metabolomics, CMF models reveal differences in metabolic responses of individuals (in a postprandial meal challenge) according to anthropometric and insulin sensitivity measures. In sensitization data analysis, CMF-based methods capture differences in temporal trajectories of children according to delivery/birth mode. We demonstrate the reliability of extracted patterns using reproducibility and replicability. This is joint work with David Horner, Rasmus Bro, Ann-Marie M Schoos, Morten Arendt Rasmussen and Evrim Acar.
Contributed TalkUncovering evolving patterns in temporal data with tensor decompositions
June 2025•TRICAP (Three-way Methods in Chemistry and Psychology) 2025•Ålesund, Norway
In this talk, we discuss why existing works might be unsuitable for the task of analyzing time-evolving data and introduce two time-aware methods: t(emporal)PARAFAC and d(ynamical)CMF. With extensive synthetic experiments we compare these methods with the state-of-the-art for the task of uncovering evolving patterns in terms of accuracy, while also highlighting the benefits and limitations with respect to the three essential requirements of analyzing temporal data: (a) time-awareness, (b) structural flexibility, and (c) uniqueness. This is joint work with Carla Schenker, Max Pfeffer, Pedro Lind, Jérémy E. Cohen and Evrim Acar.
Contributed TalkdCMF: Learning interpretable evolving patterns from temporal multiway data
May 2025•Hidden structures in dynamical systems, optimization, and machine learning, Gran Sasso Science Institute•L'Aquila, Italy
In this work, we bridge the gap between tensor factorizations and dynamical modeling and propose d(ynamical)CMF. dCMF constrains the temporal evolution of the latent factors to adhere to a specific LDS structure, thus taking into account the order of the observations. We explore how the proposed method is related to CP, PARAFAC2 and t(emporal)PARAFAC2. We highlight also the fact that if an estimate of the transition matrix is known a-priori, the framework allows for promoting this specific structure on the evolving factors. This is joint work with Carla Schenker, Jérémy E. Cohen and Evrim Acar.
Contributed TalkA Time-aware tensor decomposition for concept evolution
March 2024•94th Annual meeting of the International Association of Applied Mathematics and Mechanics (GAMM 2024)•Magdeburg, Germany
In this talk, we introduce the temporal PARAFAC2 (tPARAFAC2) model, a PARAFAC2-based tensor factorization with temporal regularization to compute a time-aware factorization of the input with the goal of extracting gradually evolving patterns, which is essential for understanding the data’s continuous development through the underlying temporal dynamics. We use an Alternating Optimization (AO) - Alternating Direction Method of Multipliers (ADMM)-based algorithm to fit the model and study different algorithmic approaches to handle missing data when fitting the model. Using numerical experiments on simulated and real data, we demonstrate the effectiveness of tPARAFAC2 model in terms of recovering the underlying (evolving) patterns accurately in various challenging cases, in particular, in the presence of missing entries. This is joint work with Max Pfeffer, Pedro Lind and Evrim Acar.
Academic Service
Reviewing
ICML 2026 workshop: Connecting Low-rank Representations in AI (CoLorAI'26)
Teaching
Spring 2026Data Mining at Scale: Algorithms and Systems (Lecturer), OsloMet
Spring 2021Data mining techniques (Teaching assistant), National and Kapodistrian University of Athens
Winter 2021Numerical analysis (Teaching assistant), National and Kapodistrian University of Athens
Winter 2020Numerical analysis (Teaching assistant), National and Kapodistrian University of Athens
News
October 2026I successfully defended my PhD thesis, titled Time-aware methods for analyzing temporal data. I also presented my trial lecture on State-space methods and sequential latent variable methods for analyzing dynamical systems.
July 2026Our work Extracting host-specific developmental signatures from longitudinal microbiome data is now published in PLOS Computational Biology. 📄 Paper
July 2026I presented my work Revealing Subject-Specific Temporal Patterns from Longitudinal Data in the Greeks in AI conference. 🌐 Website
February 2026A new preprint is available Revealing Subject-Specific Temporal Patterns from Longitudinal Data. 📄 Paper
November 2025A new preprint is available Extracting host-specific developmental signatures from longitudinal microbiome data. 📄 Paper
September 2025Our work tPARAFAC2: Tracking evolving patterns in (incomplete) temporal data is now published! 📄 Paper
September 2025Our work dCMF: Learning interpretable evolving patterns from temporal multiway data is now published! 📄 Paper
June 2025I presented my talk Uncovering evolving patterns in temporal data with tensor decompositions at TRICAP (Three-way Methods in Chemistry and Psychology) 2025 workshop at Ålesund, Norway.
May 2025I presented my talk dCMF: Learning interpretable evolving patterns from temporal multiway data at Hidden structures in dynamical systems, optimization, and machine learning workshop at Gran Sasso Science Institute. Exciting! 🌐 Website
February 2025A new preprint is available: dCMF: Learning interpretable evolving patterns from temporal multiway data. To appear in EUSIPCO 2025. 📄 Paper💻 Code
July 2024A new preprint is available: tPARAFAC2: Tracking evolving patterns in (incomplete) temporal data. To appear in Data Mining and Knowledge Discovery. 📄 Paper💻 Code
January 2024I presented my poster A time-aware tensor decomposition for concept evolution at Geilo Winter School in (you guessed it) Geilo, Norway.
October 2023Our work A Time-Aware Tensor Decomposition for Tracking Evolving Patterns is now published. 📄 Paper