Research
Research behind the tools.
plotly-resampler, tsflex and tsdownsample came out of peer-reviewed work at Ghent University (imec, predict-idlab); FlexViz builds on it.
Jeroen Van Der Donckt is a PhD in Computer Science Engineering, (co-)authored the papers and the libraries. 12 peer-reviewed publications in total, 9 as shared first author, plus a filed US patent.
Selected publications.
Time-series visualization
2022
Plotly-Resampler: Effective Visual Analytics for Large Time Series
Dynamic, view-aware aggregation that keeps large Plotly line charts responsive on every zoom and pan.
(short paper)
MinMaxLTTB: Leveraging MinMax-Preselection to Scale LTTB
A MinMax preselection step that scales Largest-Triangle-Three-Buckets downsampling to much larger series at comparable visual quality.
2023
tsdownsample: High-performance time series downsampling for scalable visualization
A Rust-accelerated downsampling library with SIMD-aware, multithreaded implementations of MinMaxLTTB, MinMax, M4 and LTTB.
Time-series processing
2022
tsflex: Flexible Time Series Processing & Feature Extraction
A toolkit for processing and extracting features from irregular, asynchronous time series across multiple window and stride configurations.
Applied machine learning
Processing and
Control, 2023
Do Not Sleep on Traditional Machine Learning: Simple and Interpretable Techniques Are Competitive to Deep Learning for Sleep Scoring
Shows that carefully engineered, interpretable models match deep learning on sleep-stage classification.
2024 (HASCA)
Left-Right Swapping and Upper-Lower Limb Pairing for Robust Multi-Wearable Workout Activity Detection
Two data augmentation techniques for multi-wearable activity recognition that correct for inconsistent device orientation.
2024 (HASCA)
Magnitude and Rotation Invariant Detection of Transportation Modes with Missing Data Modalities
A rotation-invariant feature aggregation method for transportation mode detection on smartphone data with missing modalities.
Medical prediction
2025
Improved IOL Power Calculation With Femtosecond Laser Enhanced Refractive Outcome Prediction
A clinical prediction algorithm for intraocular lens power calculation using pre- and intraoperative measurements.
2024
Mitigating Data Quality Challenges in Ambulatory Wrist-Worn Wearable Monitoring Through Analytical and Practical Approaches
Analytical and practical methods for handling the data-quality problems that arise in long-term ambulatory wearable monitoring.
Feature selection and industrial monitoring
2022
Powershap: A Power-full Shapley Feature Selection Method
A feature selection method that uses statistical hypothesis testing and power calculations on Shapley values.
Technology, 2025
Impact of Window Sizes and Sensor Quality on MCSA for Misalignment Fault Detection
How analysis window size and sensor quality affect motor current signature analysis for detecting misalignment faults on a production line.
Competitions.
challenge
UbiComp/ISWC 2024 (HASCA), 1st place
Multi-wearable workout activity detection using left-right swapping and upper-lower limb pairing, with tsflex for feature extraction.
challenge
UbiComp/ISWC 2024 (HASCA), 1st place
Transportation mode detection on smartphone data with missing modalities, with tsflex for feature extraction.
Kaggle Tabular Playground Series, 3rd of 816 teams
A sensor time-series classification task, using tsflex for feature extraction.
Kaggle BirdCLEF 2022, 11th of 807 teams
Bird call identification from soundscape recordings. Gold medal.
The software that came out of this work is on the open-source page.