2024

Differentially-Private Federated Learning with Non-IID Data For Surgical Risk Prediction. Pfitzner, Bjarne; Maurer, Max M.; Winter, Axel; Riepe, Christoph; Sauer, Igor M.; van de Water, Robin; Arnrich, Bert in 2024 IEEE First International Conference on Artificial Intelligence for Medicine, Health and Care (AIMHC) (2024). 120–129.
[ Abstract ] [ BibTeX ] [ URL ] [ DOI ]
 
Federated learning (FL) has emerged as a promising solution to deal with the privacy concerns which often limit access to data for training machine learning (ML) models. It encompasses the exchange of model parameters between data owners and a central model aggregation instance without the need to share sensitive data directly. Although FL does not require sharing sensitive information, it may not guarantee complete privacy, as model parameters can potentially reveal sensitive details about the underlying training data. To enhance privacy, FL approaches can additionally integrate differential privacy (DP), which entails the careful addition of noise to the model. In this paper, we investigate the relationship between FL, DP, and highly non-independent and identically distributed (non-IID) data. We present a specific real-life healthcare applica tion of predicting patient mortality and revision surgery after visceral operations. Our contributions encompass assessing the performance of FL with and without DP in the presence of highly non-IID data, analysing the fairness for individual training participants and investigating the efficacy of personalisation to alleviate the privacy-utility trade-off introduced by DP. Our findings indicate that the addition of DP into FL given non-IID datasets disparately favours participants with larger datasets, while the model has limited utility for those with less data. The addition of a subsequent model fine-tuning step can remedy this disparity and enables all participants to reach a model performance comparable to the centralised training scenario.
@article{pfitzner2024differentially, abstract = {Federated learning (FL) has emerged as a promising solution to deal with the privacy concerns which often limit access to data for training machine learning (ML) models. It encompasses the exchange of model parameters between data owners and a central model aggregation instance without the need to share sensitive data directly. Although FL does not require sharing sensitive information, it may not guarantee complete privacy, as model parameters can potentially reveal sensitive details about the underlying training data. To enhance privacy, FL approaches can additionally integrate differential privacy (DP), which entails the careful addition of noise to the model. In this paper, we investigate the relationship between FL, DP, and highly non-independent and identically distributed (non-IID) data. We present a specific real-life healthcare applica tion of predicting patient mortality and revision surgery after visceral operations. Our contributions encompass assessing the performance of FL with and without DP in the presence of highly non-IID data, analysing the fairness for individual training participants and investigating the efficacy of personalisation to alleviate the privacy-utility trade-off introduced by DP. Our findings indicate that the addition of DP into FL given non-IID datasets disparately favours participants with larger datasets, while the model has limited utility for those with less data. The addition of a subsequent model fine-tuning step can remedy this disparity and enables all participants to reach a model performance comparable to the centralised training scenario.}, author = {Pfitzner, Bjarne and Maurer, Max M. and Winter, Axel and Riepe, Christoph and Sauer, Igor M. and van de Water, Robin and Arnrich, Bert}, journal = {2024 IEEE First International Conference on Artificial Intelligence for Medicine, Health and Care (AIMHC)}, keywords = {sys:relevantfor:dhc bertarnrich bjarnepfitzner myown robinvandewater}, pages = {120-129}, title = {Differentially-Private Federated Learning with Non-IID Data For Surgical Risk Prediction}, year = 2024 }

2023

HARE: Unifying the Human Activity Recognition Engineering Workflow. Konak, Orhan; van de Water, Robin; Döring, Valentin; Fiedler, Tobias; Liebe, Lucas; Masopust, Leander; Postnov, Kirill; Sauerwald, Franz; Treykorn, Felix; Wischmann, Alexander; Gjoreski, Hristijan; Luštrek, Mitja; Arnrich, Bert in Sensors (2023). 23(23)
[ BibTeX ] [ URL ] [ DOI ]
 
@article{s23239571, author = {Konak, Orhan and van de Water, Robin and Döring, Valentin and Fiedler, Tobias and Liebe, Lucas and Masopust, Leander and Postnov, Kirill and Sauerwald, Franz and Treykorn, Felix and Wischmann, Alexander and Gjoreski, Hristijan and Luštrek, Mitja and Arnrich, Bert}, journal = {Sensors}, keywords = {bertarnrich orhankonak robinvandewater widehealth}, number = 23, title = {HARE: Unifying the Human Activity Recognition Engineering Workflow}, volume = 23, year = 2023 }

A Real-time Human Pose Estimation Approach for Optimal Sensor Placement in Sensor-based Human Activity Recognition. Konak, Orhan; Wischmann, Alexander; van De Water, Robin; Arnrich, Bert (2023). 1–6.
[ BibTeX ]
 
@inproceedings{konak2023real, author = {Konak, Orhan and Wischmann, Alexander and van De Water, Robin and Arnrich, Bert}, booktitle = {Proceedings of the 8th international Workshop on Sensor-Based Activity Recognition and Artificial Intelligence}, keywords = {bertarnrich orhankonak robinvandewater}, pages = {1–6}, title = {A Real-time Human Pose Estimation Approach for Optimal Sensor Placement in Sensor-based Human Activity Recognition}, year = 2023 }

2022

Farm Your ML-based Query Optimizer’s Food - Human-Guided Training Data Generation. van de Water, Robin; Ventura, Francesco; Kaoudi, Z; Quiane-Ruiz, J; Markl, Volker (2022).
[ BibTeX ] [ URL ]
 
@inproceedings{vandewaterFarmYourMLbased2022, author = {{van de Water}, Robin and Ventura, Francesco and Kaoudi, Z and {Quiane-Ruiz}, J and Markl, Volker}, booktitle = {Conference on {{Innovative Data Systems Research}} ({{CIDR}})}, keywords = {prior-hpi robinvandewater}, title = {Farm Your ML-based Query Optimizer's Food - Human-Guided Training Data Generation}, year = 2022 }

Farming Your ML-based Query Optimizer’s Food. Van De Water, Robin; Ventura, Francesco; Kaoudi, Zoi; Quiané-Ruiz, Jorge-Arnulfo; Markl, Volker (2022). 3186–3189.
[ Abstract ] [ BibTeX ] [ URL ]
 
Machine learning (ML) is becoming a core component in query optimizers, e.g., to estimate costs or cardinalities. This means large heterogeneous sets of labeled query plans or jobs (i.e., plans with their runtime or cardinality output) are needed. However, collecting such a training dataset is a very tedious and time-consuming task: It requires both developing numerous jobs and executing them to acquire ground-truth labels. We demonstrate Datafarm,a novel framework for efficiently generating and labeling training data for ML-based query optimizers to overcome these issues. Datafarmenables generating training data tailored to users' needs by learning from their existing workload patterns, input data, and computational resources. It uses an active learning approach to determine a subset of jobs to be executed and encloses the human into the loop, resulting in higher quality data. The graphical user interface of Datafarmallows users to get informative details of the generated jobs and guides them through the generation process step-by-step. We show how users can intervene and provide feedback to the system in an iterative fashion. As an output, users can download both the generated jobs to use as a benchmark and the training data (jobs with their labels).
@inproceedings{vandewaterFarmingYourMLbased2022, abstract = {Machine learning (ML) is becoming a core component in query optimizers, e.g., to estimate costs or cardinalities. This means large heterogeneous sets of labeled query plans or jobs (i.e., plans with their runtime or cardinality output) are needed. However, collecting such a training dataset is a very tedious and time-consuming task: It requires both developing numerous jobs and executing them to acquire ground-truth labels. We demonstrate Datafarm,a novel framework for efficiently generating and labeling training data for ML-based query optimizers to overcome these issues. Datafarmenables generating training data tailored to users' needs by learning from their existing workload patterns, input data, and computational resources. It uses an active learning approach to determine a subset of jobs to be executed and encloses the human into the loop, resulting in higher quality data. The graphical user interface of Datafarmallows users to get informative details of the generated jobs and guides them through the generation process step-by-step. We show how users can intervene and provide feedback to the system in an iterative fashion. As an output, users can download both the generated jobs to use as a benchmark and the training data (jobs with their labels).}, author = {Van De Water, Robin and Ventura, Francesco and Kaoudi, Zoi and Quiané-Ruiz, Jorge-Arnulfo and Markl, Volker}, keywords = {prior-hpi robinvandewater}, organization = {IEEE}, pages = {3186–3189}, title = {Farming Your ML-based Query Optimizer's Food}, year = 2022 }

2020

The Care2Report System: Automated Medical Reporting as an Integrated Solution to Reduce Administrative Burden in Healthcare. Maas, Lientje; Geurtsen, Mathan; Nouwt, Florian; Schouten, Stefan F; Van De Water, Robin; Van Dulmen, Sandra; Dalpiaz, Fabiano; Van Deemter, Kees; Brinkkemper, Sjaak in Information Technology in Healthcare: IT Architectures and Implementations in Healthcare Environments (2020).
[ Abstract ] [ BibTeX ] [ URL ]
 
Documenting patient medical information in the electronic medical record is a time-consuming task at the expense of direct patient care. We propose an integrated solution to automate the process of medical reporting. This vision is enabled through the integration of speech and action recognition technology with semantic interpretation based on knowledge graphs. This paper presents our dialogue summarization pipeline that transforms speech into a medical report via transcription and formal representation. We discuss the functional and technical architecture of our Care2Report system along with an initial system evaluation with data of real consultation sessions.
@incollection{maas2020care2report, abstract = {Documenting patient medical information in the electronic medical record is a time-consuming task at the expense of direct patient care. We propose an integrated solution to automate the process of medical reporting. This vision is enabled through the integration of speech and action recognition technology with semantic interpretation based on knowledge graphs. This paper presents our dialogue summarization pipeline that transforms speech into a medical report via transcription and formal representation. We discuss the functional and technical architecture of our Care2Report system along with an initial system evaluation with data of real consultation sessions.}, author = {Maas, Lientje and Geurtsen, Mathan and Nouwt, Florian and Schouten, Stefan F and Van De Water, Robin and Van Dulmen, Sandra and Dalpiaz, Fabiano and Van Deemter, Kees and Brinkkemper, Sjaak}, booktitle = {Information Technology in Healthcare: IT Architectures and Implementations in Healthcare Environments}, keywords = {prior-hpi robinvandewater}, publisher = {Hawaii International Conference on System Sciences (HICSS)}, title = {The Care2Report System: Automated Medical Reporting as an Integrated Solution to Reduce Administrative Burden in Healthcare}, year = 2020 }