Dafne van Kuppevelt
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Dafne van Kuppevelt has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Emla Fitzsimons
Centre for Longitudinal Studies, UCL Institute of Education, University College London, London, UK
2 shared publications Jisk Attema
2 shared publications- Mark Hamer
Institute of Sport, Exercise and Health, University College London, London, UK
2 shared publications - Séverine Sabia
Université Paris Cité, INSERM U1153, Epidemiology of Ageing and Neurodegenerative Diseases (EpiAgeing), France
2 shared publications - Vincent Theodoor van Hees
Accelting, Almere, The Netherlands
2 shared publications - A. Melle Lyklema
Faculty of Humanities, Department of Philosophy and Religious Studies, Utrecht University, The Netherlands
1 shared publication
Latest Publications
- Open Access
- 2 January 2026
- Published by The Open Journal in Journal of Open Source Education
- Vol. 9 (95)
Introduction to deep learning: Carpentries-style hands-on lesson material for introducing researchers to deep learning
This article describes a hands-on introduction to the first steps in deep learning, intended for researchers who are familiar with (non-deep) machine learning. The use of deep learning has seen a sharp increase in popularity and applicability over the last decade. While deep learning can be a useful tool for researchers from a wide range of domains, taking the first steps in the world of deep learning can be somewhat intimidating. This introduction aims to cover the fundamentals of deep learning in a practical and hands-on manner. By the end of the course, students will be able to train their first neural network and understand the subsequent steps needed to improve the model. The lesson starts by explaining the basic concepts of neural networks, and then guides learners through the different steps of a deep learning workflow. After following this lesson, learners will be able to prepare data for deep learning, implement a basic deep learning model in Python with Keras, and monitor and troubleshoot the training process. In addition, they will be able to implement and understand different layer types, such as convolutional layers and dropout layers, and apply transfer learning. We use data with permissive licenses and designed for real world use cases: • The Penguin dataset (Horst et al. (2020)) • The Weather prediction dataset (Huber et al. (2022)) • The Dollar Street Dataset (Gaviria Rojas et al. (2022)) is representative and contains accurate demographic information to ensure their robustness and fairness, especially for smaller subpopulations. Statement of Need This lesson addresses the need for an introductory lesson on deep learning that is opensource, and can be used by instructors in a workshop as well as for self-study. While generally usable, its target audience are academic researchers.
- 2 January 2026
- Published by The Open Journal in Journal of Open Source Education
- Vol. 9 (95)
Open AccessIntroduction to deep learning: Carpentries-style hands-on lesson material for introducing researchers to deep learning
This article describes a hands-on introduction to the first steps in deep learning, intended for researchers who are familiar with (non-deep) machine learning. The use of deep learning has seen a sharp increase in popularity and applicability over the last decade. While deep learning can be a useful tool for researchers from a wide range of domains, taking the first steps in the world of deep learning can be somewhat intimidating. This introduction aims to cover the fundamentals of deep learning in a practical and hands-on manner. By the end of the course, students will be able to train their first neural network and understand the subsequent steps needed to improve the model. The lesson starts by explaining the basic concepts of neural networks, and then guides learners through the different steps of a deep learning workflow. After following this lesson, learners will be able to prepare data for deep learning, implement a basic deep learning model in Python with Keras, and monitor and troubleshoot the training process. In addition, they will be able to implement and understand different layer types, such as convolutional layers and dropout layers, and apply transfer learning. We use data with permissive licenses and designed for real world use cases: • The Penguin dataset (Horst et al. (2020)) • The Weather prediction dataset (Huber et al. (2022)) • The Dollar Street Dataset (Gaviria Rojas et al. (2022)) is representative and contains accurate demographic information to ensure their robustness and fairness, especially for smaller subpopulations. Statement of Need This lesson addresses the need for an introductory lesson on deep learning that is opensource, and can be used by instructors in a workshop as well as for self-study. While generally usable, its target audience are academic researchers.
- Open Access
- 19 November 2021
- Published by Springer Nature in Journal of Computational Social Science
- Vol. 5 (1) , 841-860
Community membership consistency applied to corporate board interlock networks
Community detection is a well-established method for studying the meso-scale structure of social networks. Applying a community detection algorithm results in a division of a network into communities that is often used to inspect and reason about community membership of specific nodes. This micro-level interpretation step of community structure is a crucial step in typical social science research. However, the methodological caveat in this step is that virtually all modern community detection methods are non-deterministic and based on randomization and approximated results. This needs to be explicitly taken into consideration when reasoning about community membership of individual nodes. To do so, we propose a metric of community membership consistency, that provides node-level insights in how reliable the placement of that node into a community really is. In addition, it enables us to distinguish the community core members of a community. The usefulness of the proposed metrics is demonstrated on corporate board interlock networks, in which weighted links represent shared senior level directors between firms. Results suggest that the community structure of global business groups is centered around persistent communities consisting of core countries tied by geographical and cultural proximity. In addition, we identify fringe countries that appear to associate with a number of different global business communities.
- 19 November 2021
- Published by Springer Nature in Journal of Computational Social Science
- Vol. 5 (1) , 841-860
Open AccessCommunity membership consistency applied to corporate board interlock networks
Community detection is a well-established method for studying the meso-scale structure of social networks. Applying a community detection algorithm results in a division of a network into communities that is often used to inspect and reason about community membership of specific nodes. This micro-level interpretation step of community structure is a crucial step in typical social science research. However, the methodological caveat in this step is that virtually all modern community detection methods are non-deterministic and based on randomization and approximated results. This needs to be explicitly taken into consideration when reasoning about community membership of individual nodes. To do so, we propose a metric of community membership consistency, that provides node-level insights in how reliable the placement of that node into a community really is. In addition, it enables us to distinguish the community core members of a community. The usefulness of the proposed metrics is demonstrated on corporate board interlock networks, in which weighted links represent shared senior level directors between firms. Results suggest that the community structure of global business groups is centered around persistent communities consisting of core countries tied by geographical and cultural proximity. In addition, we identify fringe countries that appear to associate with a number of different global business communities.
- 2 August 2021
stroll-srl
- 2 August 2021
stroll-srl
- Open Access
- 20 July 2021
- Published by Walter de Gruyter GmbH in Islamic Law and Society
- Vol. 28 (3) , 234-281
Text Mining Islamic Law
Digital humanities has a venerable pedigree, stretching back to the middle of the twentieth century, but despite noteworthy pioneering contributions it has not become a mainstream practice in Islamic Studies. This essay applies humanities computing to the study of Islamic law. We analyze a representative corpus of works of Islamic substantive law (furūʿ al-fiqh) from the beginnings of Islamic legal jurisprudence to the early modern period (2nd/8th-13th/19th c.) using several computational tools and methods: text-reuse network analysis based on plain-text annotations and html tags, clustered frequency-based analysis, word clouds, and topic modeling. Applying machine-guided distant reading to Islamic legal texts over the longue-dureé, we study (1) the role of the Qurʾān, (2) patterns of normative qualifications (aḥkām), and (3) the distribution of topics in our corpus. In certain instances the analysis confirms claims made in the scholarly literature on Islamic law, in other instances it corrects such claims.
- 20 July 2021
- Published by Walter de Gruyter GmbH in Islamic Law and Society
- Vol. 28 (3) , 234-281
Open AccessText Mining Islamic Law
Digital humanities has a venerable pedigree, stretching back to the middle of the twentieth century, but despite noteworthy pioneering contributions it has not become a mainstream practice in Islamic Studies. This essay applies humanities computing to the study of Islamic law. We analyze a representative corpus of works of Islamic substantive law (furūʿ al-fiqh) from the beginnings of Islamic legal jurisprudence to the early modern period (2nd/8th-13th/19th c.) using several computational tools and methods: text-reuse network analysis based on plain-text annotations and html tags, clustered frequency-based analysis, word clouds, and topic modeling. Applying machine-guided distant reading to Islamic legal texts over the longue-dureé, we study (1) the role of the Qurʾān, (2) patterns of normative qualifications (aḥkām), and (3) the distribution of topics in our corpus. In certain instances the analysis confirms claims made in the scholarly literature on Islamic law, in other instances it corrects such claims.
- Open Access
- 9 January 2019
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 14 (1) , e0208692
Segmenting accelerometer data from daily life with unsupervised machine learning
Accelerometers are increasingly used to obtain valuable descriptors of physical activity for health research. The cut-points approach to segment accelerometer data is widely used in physical activity research but requires resource expensive calibration studies and does not make it easy to explore the information that can be gained for a variety of raw data metrics. To address these limitations, we present a data-driven approach for segmenting and clustering the accelerometer data using unsupervised machine learning. The data used came from five hundred fourteen-year-old participants from the Millennium cohort study who wore an accelerometer (GENEActiv) on their wrist on one weekday and one weekend day. A Hidden Semi-Markov Model (HSMM), configured to identify a maximum of ten behavioral states from five second averaged acceleration with and without addition of x, y, and z-angles, was used for segmenting and clustering of the data. A cut-points approach was used as comparison. Time spent in behavioral states with or without angle metrics constituted eight and five principal components to reach 95% explained variance, respectively; in comparison four components were identified with the cut-points approach. In the HSMM with acceleration and angle as input, the distributions for acceleration in the states showed similar groupings as the cut-points categories, while more variety was seen in the distribution of angles. Our unsupervised classification approach learns a construct of human behavior based on the data it observes, without the need for resource expensive calibration studies, has the ability to combine multiple data metrics, and offers a higher dimensional description of physical behavior. States are interpretable from the distributions of observations and by their duration.
- 9 January 2019
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 14 (1) , e0208692
Open AccessSegmenting accelerometer data from daily life with unsupervised machine learning
Accelerometers are increasingly used to obtain valuable descriptors of physical activity for health research. The cut-points approach to segment accelerometer data is widely used in physical activity research but requires resource expensive calibration studies and does not make it easy to explore the information that can be gained for a variety of raw data metrics. To address these limitations, we present a data-driven approach for segmenting and clustering the accelerometer data using unsupervised machine learning. The data used came from five hundred fourteen-year-old participants from the Millennium cohort study who wore an accelerometer (GENEActiv) on their wrist on one weekday and one weekend day. A Hidden Semi-Markov Model (HSMM), configured to identify a maximum of ten behavioral states from five second averaged acceleration with and without addition of x, y, and z-angles, was used for segmenting and clustering of the data. A cut-points approach was used as comparison. Time spent in behavioral states with or without angle metrics constituted eight and five principal components to reach 95% explained variance, respectively; in comparison four components were identified with the cut-points approach. In the HSMM with acceleration and angle as input, the distributions for acceleration in the states showed similar groupings as the cut-points categories, while more variety was seen in the distribution of angles. Our unsupervised classification approach learns a construct of human behavior based on the data it observes, without the need for resource expensive calibration studies, has the ability to combine multiple data metrics, and offers a higher dimensional description of physical behavior. States are interpretable from the distributions of observations and by their duration.
- Open Access
Segmenting accelerometer data from daily life with unsupervised machine learning
Purpose: Accelerometers are increasingly used to obtain valuable descriptors of physical activity for health research. The cut-points approach to segment accelerometer data is widely used in physical activity research but requires resource expensive calibration studies and does not make it easy to explore the information that can be gained for a variety of raw data metrics. To address these limitations, we present a data-driven approach for segmenting and clustering the accelerometer data using unsupervised machine learning. Methods: The data used came from five hundred fourteen-year-old participants from the Millennium cohort study who wore an accelerometer (GENEActiv) on their wrist on one weekday and one weekend day. A Hidden Semi-Markov Model (HSMM), configured to identify a maximum of ten behavioral states from five second averaged acceleration with and without addition of x, y, and z-angles, was used for segmenting and clustering of the data. A cut-points approach was used as comparison. Results: Time spent in behavioral states with or without angle metrics constituted eight and five principal components to reach 95% explained variance, respectively; in comparison four components were identified with the cut-points approach. In the HSMM with acceleration and angle as input, the distributions for acceleration in the states showed similar groupings as the cut-points categories, while more variety was seen in the distribution of angles. Conclusion: Our unsupervised classification approach learns a construct of human behavior based on the data it observes, without the need for resource expensive calibration studies, has the ability to combine multiple data metrics, and offers a higher dimensional description of physical behavior. States are interpretable from the distributions of observations and by their duration.
Open AccessSegmenting accelerometer data from daily life with unsupervised machine learning
Purpose: Accelerometers are increasingly used to obtain valuable descriptors of physical activity for health research. The cut-points approach to segment accelerometer data is widely used in physical activity research but requires resource expensive calibration studies and does not make it easy to explore the information that can be gained for a variety of raw data metrics. To address these limitations, we present a data-driven approach for segmenting and clustering the accelerometer data using unsupervised machine learning. Methods: The data used came from five hundred fourteen-year-old participants from the Millennium cohort study who wore an accelerometer (GENEActiv) on their wrist on one weekday and one weekend day. A Hidden Semi-Markov Model (HSMM), configured to identify a maximum of ten behavioral states from five second averaged acceleration with and without addition of x, y, and z-angles, was used for segmenting and clustering of the data. A cut-points approach was used as comparison. Results: Time spent in behavioral states with or without angle metrics constituted eight and five principal components to reach 95% explained variance, respectively; in comparison four components were identified with the cut-points approach. In the HSMM with acceleration and angle as input, the distributions for acceleration in the states showed similar groupings as the cut-points categories, while more variety was seen in the distribution of angles. Conclusion: Our unsupervised classification approach learns a construct of human behavior based on the data it observes, without the need for resource expensive calibration studies, has the ability to combine multiple data metrics, and offers a higher dimensional description of physical behavior. States are interpretable from the distributions of observations and by their duration.