publications

2026

1. arxiv

   Pruning Extensions and Efficiency Trade-Offs for Sustainable Time Series Classification

   Raphael Fischer, Angus Dempster, Sebastian Buschjäger, and 3 more authors

   2026

   PAPER Bibtex

   @misc{Fischer/etal/2026,
     title = {Pruning Extensions and Efficiency Trade-Offs for Sustainable Time Series Classification},
     author = {Fischer, Raphael and Dempster, Angus and Buschjäger, Sebastian and Jakobs, Matthias and Maniar, Urav and Webb, Geoffrey I.},
     year = {2026},
     eprint = {2604.07953},
     archiveprefix = {arXiv},
     primaryclass = {cs.LG},
   }

2. arxiv

   Jump Like A Squirrel: Optimized Execution Step Order for Anytime Random Forest Inference

   Daniel Biebert, Christian Hakert, Kay Heider, and 3 more authors

   2026

   PAPER Bibtex

   @misc{Biebert/etal/2026,
     title = {Jump Like A Squirrel: Optimized Execution Step Order for Anytime Random Forest Inference},
     author = {Biebert, Daniel and Hakert, Christian and Heider, Kay and Kuhse, Daniel and Buschjäger, Sebastian and Chen, Jian-Jia},
     year = {2026},
     eprint = {2603.01588},
     archiveprefix = {arXiv},
     primaryclass = {cs.LG},
   }

3. ECML Journal Track

   Lift What You Can: Green Online Learning with Heterogeneous Ensembles

   Kirsten Köbschall, Sebastian Buschjäger, Raphael Fischer, and 2 more authors

   2026

   Abs PAPER Bibtex

   Ensemble methods for stream mining necessitate managing multiple models and updating them as data distributions evolve. Considering the calls for more sustainability, established methods are however not sufficiently considerate of ensemble members’ computational expenses and instead overly focus on predictive capabilities. To address these challenges and enable green online learning, we propose heterogeneous online ensembles (HEROS). For every training step, HEROS chooses a subset of models from a pool of models initialized with diverse hyperparameter choices under resource constraints to train. We introduce a Markov decision process to theoretically capture the trade-offs between predictive performance and sustainability constraints. Based on this framework, we present different policies for choosing which models to train on incoming data. Most notably, we propose the novel Zeta-policy, which focuses on training near-optimal models at reduced costs. Using a stochastic model, we theoretically prove that our Zeta-policy achieves near optimal performance while using fewer resources compared to the best performing policy. In our experiments across 11 benchmark datasets, we find empiric evidence that our Zeta-policy is a strong contribution to the state-of-the-art, demonstrating highly accurate performance, in some cases even outperforming competitors, and simultaneously being much more resource-friendly.

   @article{Koebschall/etal/2026,
     author = {Köbschall, Kirsten and Buschjäger, Sebastian and Fischer, Raphael and Hartung, Lisa and Kramer, Stefan},
     title = {Lift What You Can: Green Online Learning with Heterogeneous Ensembles},
     journaltitle = {Data Mining and Knowledge Discovery},
     year = {2026},
     volume = {40},
     number = {3},
     eid = {32},
     doi = {10.1007/s10618-026-01200-3},
     url = {https://link.springer.com/article/10.1007/s10618-026-01200-3},
     issn = {1573-756X},
     eprint = {2509.18962},
     eprintclass = {cs.LG},
     pubstate = {inpress},
   }

2025

1. ECML WS

   Improved Sleep Stage Tagging on Wearables via Knowledge Distillation

   Hendrik Weissenfels, Matthias Jakobs, Sarah Dietz-Terjung, and 1 more author

   In 2025 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases Workshop Track (ECML PKDD Workshop Track) – Learning on Real and Synthetic Medical Time Series Data (MedTime) , 2025

   Abs Bibtex

   Sleep disorders present a major global health challenge, requiring precise and efficient diagnostic approaches. Polysomnography (PSG) remains the gold standard by capturing multiple physiological signals during overnight monitoring. However, PSG is both resource-intensive and burdensome for patients. A key limitation is the so-called first-night effect, i.e. the altered sleep architecture due to the unfamiliar laboratory environment and sensor setup. This can lead to artificially reduced sleep efficiency and diagnostic distortion. Wearable-based sleep monitoring offers a promising alternative, enabling longitudinal assessment and reducing the first-night effect by using fewer sensors and enabling home-based diagnostics. Recent advancements in wearable technologies have improved accuracy in estimating sleep parameters when compared to PSG. This study investigates knowledge distillation (KD) as a strategy to transfer knowledge from an existing high-performing PSG-based teacher model to a student model operating on wearable-like input data. Both models use the UTime architecture which has been shown to perform well across different sleep laboratories and are trained on clinical PSG data from a sleep laboratory in Germany. Our KD framework combines soft target supervision and feature matching to guide the student’s learning. The results show that KD consistently improves the performance of the student in various sensor configurations and model sizes, indicating its potential for accurate sleep staging in real-world wearable applications.

   @inproceedings{Weissenfels/etal/2025,
     title = {Improved Sleep Stage Tagging on Wearables via Knowledge Distillation},
     author = {Weissenfels, Hendrik and Jakobs, Matthias and Dietz-Terjung, Sarah and Buschjäger, Sebastian},
     year = {2025},
     booktitle = {2025 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases Workshop Track (ECML PKDD Workshop Track) -- Learning on Real and Synthetic Medical Time Series Data (MedTime) },
   }

2. arxiv

   You Only Look Once at Anytime (AnytimeYOLO): Analysis and Optimization of Early-Exits for Object-Detection

   Daniel Kuhse, Harun Teper, Sebastian Buschjäger, and 2 more authors

   In arXiv, 2025

   Abs PAPER Bibtex

   We introduce AnytimeYOLO, a family of variants of the YOLO architecture that enables anytime object detection. Our AnytimeYOLO networks allow for interruptible inference, i.e., they provide a prediction at any point in time, a property desirable for safety-critical real-time applications. We present structured explorations to modify the YOLO architecture, enabling early termination to obtain intermediate results. We focus on providing fine-grained control through high granularity of available termination points. First, we formalize Anytime Models as a special class of prediction models that offer anytime predictions. Then, we discuss a novel transposed variant of the YOLO architecture, that changes the architecture to enable better early predictions and greater freedom for the order of processing stages. Finally, we propose two optimization algorithms that, given an anytime model, can be used to determine the optimal exit execution order and the optimal subset of early-exits to select for deployment in low-resource environments. We evaluate the anytime performance and trade-offs of design choices, proposing a new anytime quality metric for this purpose. In particular, we also discuss key challenges for anytime inference that currently make its deployment costly.

   @inproceedings{Kuhse/etal/2025,
     author = {Kuhse, Daniel and Teper, Harun and Buschjäger, Sebastian and Wang, Chien-Yao and Chen, Jian-Jia},
     booktitle = {arXiv},
     date = {2025-03-21},
     title = {You Only Look Once at Anytime ({AnytimeYOLO}): Analysis and Optimization of Early-Exits for Object-Detection},
     doi = {10.48550/arXiv.2503.17497},
     eprint = {2503.17497 [cs]},
     number = {{arXiv}:2503.17497},
     publisher = {{arXiv}},
     pubstate = {prepublished},
     urldate = {2025-05-09},
     keywords = {Computer Science - Computer Vision and Pattern Recognition, Computer Science - Machine Learning},
   }

3. IEEE Internet Comp.

   Rethinking Computing Systems in the Era of Climate Crisis: A Call for a Sustainable Computing Continuum

   Ella Peltonen, Suzan Bayhan, David Bermbach, and 16 more authors

   2025

   Abs PAPER Bibtex

   The advancement and widespread adoption of computing technology has yielded services that could help mitigate the climate crisis. However, the retirement of obsolete equipment, the consumption of rare earth materials, and the escalating energy demands associated with massive data processing and cloud infrastructures have raised new environmental dilemmas. Existing design and development methodologies primarily focus on fulfilling functional requirements and improving performance. In this paper, we argue that these methodologies must be augmented with sustainability considerations encompassing energy efficiency, material usage, longevity, and upgradability. Solutions at different layers of the system stack, from the physical to the application layer, must be integrated. Moreover, there should be a strong focus on the transparency of sustainability metrics across the whole computing continuum. Building on fruitful discussions at the international Lorentz workshop on Future Computing for Digital Infrastructures, we advocate novel approaches in the design, development, and operation of the computing continuum.

   @article{Peltonen/etal/2025,
     author = {Peltonen, Ella and Bayhan, Suzan and Bermbach, David and Buschjager, Sebastian and Degeler, Victoria and Ding, Aaron Yi and Incel, Ozlem Durmaz and Katare, Dewant and Kjargaard, Mikkel Baun and Leroux, Sam and Mahmoodi, Toktam and Mann, Zoltan Adam and Meratnia, Nirvana and Pimentel, Andy D. and Rellermeyer, Jan S. and Riviere, Etienne and Sapra, Dolly and Solmaz, Gurkan and Waaij, Bram van der},
     date = {2025-05},
     journaltitle = {IEEE Internet Computing},
     title = {{ Rethinking Computing Systems in the Era of Climate Crisis: A Call for a Sustainable Computing Continuum }},
     doi = {10.1109/MIC.2025.3566642},
     issn = {1941-0131},
     number = {01},
     pages = {1--9},
     pubstate = {inpress},
     url = {https://doi.ieeecomputersociety.org/10.1109/MIC.2025.3566642},
     keywords = {Data centers;Sustainable development;Energy efficiency;Regulation;Meteorology;Energy consumption;Costs;Artificial intelligence;Streaming media;Hardware},
     location = {Los Alamitos, CA, USA},
     publisher = {IEEE Computer Society},
   }

4. MLJ

   Splitting Stump Forests: Tree Ensemble Compression for Edge Devices (extended version)

   Fouad Alkhoury, Sebastian Buschjäger, and Pascal Welke

   2025

   Abs PAPER Bibtex

   We introduce Splitting Stump Forests—small ensembles of weak learners extracted from a trained random forest. The high memory consumption of random forests renders them unfit for resource-constrained devices. We show empirically that we can significantly reduce the model size and inference time by selecting nodes that evenly split the arriving training data and applying a linear model on the resulting representation. Our extensive empirical evaluation indicates that Splitting Stump Forests outperform random forests and state-of-the-art compression methods on memory-limited embedded devices.

   @article{Alkhoury/etal/2025,
     author = {Alkhoury, Fouad and Buschjäger, Sebastian and Welke, Pascal},
     date = {2025-08-27},
     journaltitle = {Machine Learning},
     pubstate = {inpress},
     title = {Splitting Stump Forests: Tree Ensemble Compression for Edge Devices (extended version)},
     doi = {10.1007/s10994-025-06866-2},
     issn = {1573-0565},
     number = {10},
     pages = {219},
     url = {https://doi.org/10.1007/s10994-025-06866-2},
     urldate = {2026-01-05},
     volume = {114},
     keywords = {Ensemble compression, Random forests, Edge devices},
     langid = {english},
     shortjournal = {Mach Learn},
     shorttitle = {Splitting stump forests},
   }

5. Pneumologie

   Beyond Sleep Staging: Advancing End-to-End Event Scoring in Sleep Medicine

   S. Dietz-Terjung, M. Jakobs, C. Labeit, and 2 more authors

   Mar 2025

   Abs PAPER Bibtex

   Background: Sleep disorders are a global health concern, with polysomnography (PSG), the gold standard for diagnosis. Traditional PSG analysis involves manual annotation of various physiological signals, including EEG, EOG, EMG, ECG, airflow, and oximetry making the process time-consuming. Recent advances in deep learning have enabled automated sleep staging using EEG data. However, existing solutions are limited to specific data subsets. This study aims to expand deep-learning applications to utilize the full spectrum of PSG data for a comprehensive analysis of sleep-related events. Material and Methods: We developed an end-to-end machine learning system that leverages all available PSG sensor channels. Our approach scores 19 different events across six event groups, such as breathing patterns, heart rate, and movement during sleep. We explored two models: one trained separately per event group using selected PSG channels, and a joint model predicting all events simultaneously using all channels. To address overfitting issues in the joint model, we implemented curriculum learning, which introduces tasks in a structured manner to improve model generalization. Results: Systematic experimentation demonstrated that curriculum learning effectively reduced overfitting in the joint model, closing the performance gap between the joint and multiple models. Additionally, a qualitative analysis of channel importance highlighted the advantages of curriculum learning in model training. Our approach outperformed commercial systems based on hand-crafted rules, showing significant improvement in event scoring accuracy. Conclusion: This study presents the first comprehensive machine learning system utilizing full PSG data, bridging the gap between deep learning research and its practical application in clinical sleep medicine. The use of curriculum learning enhances model performance, offering a promising tool for more accurate and efficient diagnosis of sleep disorders.

   @article{DietzTerjung/etal/2025a,
     author = {Dietz-Terjung, S. and Jakobs, M. and Labeit, C. and Sch{\"o}bel, C. and Buschj{\"a}ger, S.},
     journaltitle = {Pneumologie},
     title = {Beyond Sleep Staging: Advancing End-to-End Event Scoring in Sleep Medicine},
     doi = {10.1055/s-0045-1804722},
     issn = {0934-8387},
     language = {DE},
     number = {S 01},
     pages = {P 118},
     pubstate = {inpress},
     url = {https://doi.org/10.1055/s-0045-1804722},
     volume = {79},
     day = {18},
     month = mar,
     publisher = {Georg Thieme Verlag KG},
     year = {2025}
   }

6. ECML WS

   Decentralized Time Series Classification with ROCKET Features

   Bruno Casella, Matthias Jakobs, Marco Aldinucci, and 1 more author

   In 2025 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases Workshop Track (ECML PKDD Workshop Track) – WAFL - 3rd Workshop on Advancements in Federated Learning, Mar 2025

   Abs PAPER Bibtex

   Time series classification (TSC) is a critical task with applications in various domains, including healthcare, finance, and industrial monitoring. Due to privacy concerns and data regulations, Federated Learning has emerged as a promising approach for learning from distributed time series data without centralizing raw information. However, most FL solutions rely on a client-server architecture, which introduces robustness and confidentiality risks related to the distinguished role of the server, which is a single point of failure and can observe knowledge extracted from clients. To address these challenges, we propose DROCKS, a fully decentralized FL framework for TSC that leverages ROCKET (RandOm Convolutional KErnel Transform) features. In DROCKS, the global model is trained by sequentially traversing a structured path across federation nodes, where each node refines the model and selects the most effective local kernels before passing them to the successor. Extensive experiments on the UCR archive demonstrate that DROCKS outperforms state-of-the-art client-server FL approaches while being more resilient to node failures and malicious attacks. Our code is available at \urlhttps://anonymous.4open.science/r/DROCKS-7FF3/README.md.

   @inproceedings{Casella/etal/2025a,
     title = {Decentralized Time Series Classification with ROCKET Features},
     author = {Casella, Bruno and Jakobs, Matthias and Aldinucci, Marco and Buschjäger, Sebastian},
     year = {2025},
     booktitle = {2025 European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases Workshop Track (ECML PKDD Workshop Track) -- WAFL - 3rd Workshop on Advancements in Federated Learning},
     lamarr-area = {reaml},
     url = {https://anonymous.4open.science/r/DROCKS-7FF3/README.md},
   }

7. DATE

   TrackScorer: Skyrmion Logic-in-Memory Accelerator for Tree-Based Ranking Models

   Elijah Seth Cishugi, Sebastian Buschjäger, Martijn Noorlander, and 2 more authors

   In Design, Automation & Test in Europe Conference, DATE 2025, Lyon, France, March 31 - April 2, 2025, Mar 2025

   PAPER Bibtex

   @inproceedings{Cishugi/etal/2025,
     author = {Cishugi, Elijah Seth and Buschj{\"{a}}ger, Sebastian and Noorlander, Martijn and Ottavi, Marco and Chen, Kuan{-}Hsun},
     title = {TrackScorer: Skyrmion Logic-in-Memory Accelerator for Tree-Based Ranking
                       Models},
     booktitle = {Design, Automation {\&} Test in Europe Conference, {DATE} 2025,
                       Lyon, France, March 31 - April 2, 2025},
     pages = {1--7},
     publisher = {{IEEE}},
     year = {2025},
     url = {https://doi.org/10.23919/DATE64628.2025.10992934},
     doi = {10.23919/DATE64628.2025.10992934},
     timestamp = {Sat, 31 May 2025 23:09:09 +0200},
     biburl = {https://dblp.org/rec/conf/date/CishugiBNOC25.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org},
   }

8. ESANN

   Federated Time Series Classification with ROCKET features

   Bruno Casella, Matthias Jakobs, Marco Aldinucci, and 1 more author

   In European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning, Oct 2025

   Abs PAPER Bibtex

   This paper proposes FROCKS, a federated time series classification method using ROCKET features. Our approach dynamically adapts the models features by selecting and exchanging the best-performing ROCKET kernels from a federation of clients. Specifically, the server gathers the best-performing kernels of the clients together with the associated model parameters, and it performs a weighted average if a kernel is best-performing for more than one client. We compare the proposed method with state-of-the-art approaches on the UCR archive binary classification datasets and show superior performance on most datasets.

   @inproceedings{Casella/etal/2024a,
     author = {Casella, Bruno and Jakobs, Matthias and Aldinucci, Marco and Buschjäger, Sebastian},
     booktitle = {European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning},
     year = {2025},
     month = oct,
     url = {https://doi.org/10.14428/esann/2024.ES2024-61},
     title = {Federated Time Series Classification with ROCKET features},
   }

2024

1. LCTES

   Language-Based Deployment Optimization for Random Forests (Invited Paper)

   Jannik Malcher, Daniel Biebert, Kuan-Hsun Chen, and 3 more authors

   In ACM SIGPLAN/SIGBED International Conference on Languages, Compilers, and Tools for Embedded Systems, Oct 2025

   Abs PAPER Bibtex

   Arising popularity for resource-efficient machine learning models makes random forests and decision trees famous models in recent years. Naturally, these models are tuned, optimized, and transformed to feature maximally low-resource consumption. A subset of these strategies targets the model structure and model logic and therefore induces a trade-off between resource-efficiency and prediction performance. An orthogonal set of approaches targets hardware-specific optimizations, which can improve performance without changing the behavior of the model. Since such hardware-specific optimizations are usually hardware-dependent and inflexible in their realizations, this paper envisions a more general application of such optimization strategies at the level of programming languages. We therefore discuss a set of suitable optimization strategies first in general and envision their application in LLVM IR, i.e. a flexible and hardware-independent ecosystem.

   @inproceedings{Malcher/etal/2024a,
     author = {Malcher, Jannik and Biebert, Daniel and Chen, Kuan-Hsun and Buschjäger, Sebastian and Hakert, Christian and Chen, Jian-Jia},
     booktitle = {ACM SIGPLAN/SIGBED International Conference on Languages, Compilers, and Tools for Embedded Systems},
     date = {2024-06},
     url = {https://dl.acm.org/doi/10.1145/3652032.3659366},
     title = {Language-Based Deployment Optimization for Random Forests (Invited Paper)},
   }

2. ECML

   Rejection Ensembles with Online Calibration (RewOC)

   Sebastian Buschjäger

   In European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD, Oct 2024

   Abs PAPER Code Bibtex

   As machine learning models become increasingly integrated into various applications, the need for resource-aware deployment strategies becomes paramount. One promising approach for optimizing resource consumption is rejection ensembles. Rejection ensembles combine a small model deployed to an edge device with a large model deployed in the cloud with a rejector tasked to determine the most suitable model for a given input. Due to its novelty, existing research predominantly focuses on ad-hoc ensemble design, lacking a thorough understanding of rejector optimization and deployment strategies. This paper addresses this research gap by presenting a theoretical investigation into rejection ensembles and proposing a novel algorithm for training and deploying rejectors based on these novel insights. We give precise conditions of when a good rejector can improve the ensemble’s overall performance beyond the big model’s performance and when a bad rejector can make the ensemble worse than the small model. Second, we show that even the perfect rejector can overuse its budget for using the big model during deployment. Based on these insights, we propose to ignore any budget constraints during training but introduce additional safeguards during deployment. Experimental evaluation on 8 different datasets from various domains demonstrates the efficacy of our novel rejection ensembles outperforming existing approaches. Moreover, compared to standalone large model inference, we highlight the energy efficiency gains during deployment on a Nvidia Jetson AGX board.

   @inproceedings{Buschjaeger/2024,
     author = {Buschjäger, Sebastian},
     booktitle = {European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD},
     title = {Rejection Ensembles with Online Calibration (RewOC)},
     year = {2024},
     url = {https://link.springer.com/chapter/10.1007/978-3-031-70365-2_1},
   }

3. ECML

   MetaQuRe: Meta-Learning from Model Quality and Resource Consumption

   Raphael Fischer, Marcel Wever, and Sebastian Buschjäger

   In European Conference on Machine Learning and Knowledge Discovery in Databases, Oct 2024

   Abs PAPER Bibtex

   Automated machine learning (AutoML) allows for selecting, parametrizing, and composing learning algorithms for a given data set. While resources play a pivotal role in neural architecture search, it is less pronounced by classical AutoML approaches. In fact, they generally focus on only maximizing predictive quality and disregard the importance of finding resource-efficient solutions. To push resource-awareness further, our work explicitly explores how measures such as running time or energy consumption can be better considered in AutoML. Firstly, we propose a novel method for algorithm selection that balances multiple performance aspects (including resource demand) as prioritized by the user with the help of compositional meta-learning. Secondly, to foster research on green meta-learning and AutoML, we release the MetaQuRe data set, which contains information on predictive (Qu)ality and (Re)source consumption of models evaluated across hundreds of data sets and four execution environments. We use this data to put our methodology into practice and conduct an in-depth analysis of how our approach and data set can help in making AutoML more resource-aware, which represents our third contribution. Lastly, we publish MetaQuRe alongside an extensive code base, allowing for reproducing all results, expanding our data with results from custom environments, and exploring MetaQuRe interactively. In short, our work demonstrates both the importance as well as benefits of rethinking AutoML and meta-learning in a resource-aware way, thus paving the path for making future ML solutions more sustainable.

   @inproceedings{Fischer/etal/2024b,
     title = {{MetaQuRe}: Meta-Learning from Model Quality and Resource Consumption},
     author = {Fischer, Raphael and Wever, Marcel and Buschjäger, Sebastian},
     booktitle = {European Conference on Machine Learning and Knowledge Discovery in Databases},
     year = {2024},
     url = {https://link.springer.com/chapter/10.1007/978-3-031-70368-3_13}
   }

4. SEC

   Stress-Testing USB Accelerators for Efficient Edge Inference (to appear)

   Raphael Fischer, Alexander Staay, and Sebastian Buschjäger

   In ACM/IEEE Symposium on Edge Computing, Oct 2024

   Abs PAPER Bibtex

   Several manufacturers sell specialized USB devices for accelerating machine learning (ML) on the edge. While being generally promoted as a versatile solution for more efficient edge inference with deep learning models, extensive practical insights on their usability and performance are hard to find. In order to make ML deployment on the edge more sustainable, our work investigates how resource efficient these USB accelerators really are. For that, we first introduce a novel and theoretically sound methodology. It allows for comparing intricate model performance in terms of quality and resource consumption across different execution environments. We then put it into practice by studying the usability and efficiency of Google’s Coral edge tensor processing unit (TPU) and Intel’s neural compute stick 2 (NCS). In total, we benchmark over 30 models across nine hardware configurations, which reveals intricate trade-offs. Our work demonstrates that USB accelerators are indeed capable of reducing the energy consumption by a factor up to ten, however this improvement cannot be observed for all configurations - more than 50% of the investigated models cannot be run on accelerator hardware, and in several other cases, the power draw is only marginally improved. Our experiments show that the NCS improves efficiency in a more stable way, while the TPU shows further benefits in specific cases but performs less predictable. We hope that our paper provides valuable insights for practitioners that want to deploy ML on the edge in the most efficient and sustainable way.

   @inproceedings{Fischer/etal/2024,
     author = {Fischer, Raphael and van der Staay, Alexander and Buschjäger, Sebastian},
     title = {Stress-Testing USB Accelerators for Efficient Edge Inference (to appear)},
     doi = {https://doi.org/10.21203/rs.3.rs-3793927/v1},
     url = {https://www.researchsquare.com/article/rs-3793927/v1},
     booktitle = {ACM/IEEE Symposium on Edge Computing},
     year = {2024},
   }

5. MIDDLEWARE

   STRATA: Random Forests going Serverless (to appear)

   Dimitrios Tomaras, Sebastian Buschjäger, Vana Kalogeraki, and 2 more authors

   In 25th ACM/IFIP International Middleware Conference, Oct 2024

   Abs Code Bibtex

   Serverless computing has received growing interest in recent years for supporting machine learning tasks. This computational model has desirable advantages as it allows for parallelism of training tasks, exploiting the undoubtedly seamless mechanism for scaling and elastic usage of resources based on the applications’ demands, and improves manageability without the need to know the internals of the underlying technology. Training a machine learning model on top of a serverless environment is a nontrivial procedure since several challenges must be addressed, such as the communication cost of the training data, the communication patterns, the training time, and the cost of execution. In this work, we focus on Random Forests, a state-of-the-art technique in many machine learning applications. We propose STRATA, a cost-effective framework to train Random Forests on top of a serverless environment that addresses the aforementioned training challenges practically and efficiently by at least 57% on average, as we illustrate in our extensive experimental evaluation

   @inproceedings{Tomaras/etal/2024,
     author = {Tomaras, Dimitrios and Buschjäger, Sebastian and Kalogeraki, Vana and Morik, Katharina and Gunopulos, Dimitrios},
     booktitle = {25th ACM/IFIP International Middleware Conference},
     pubstate = {forthcoming},
     title = {STRATA: Random Forests going Serverless (to appear)},
     year = {2024},
   }

2023

1. ECML

   Joint leaf-refinement and ensemble pruning through L\(_\mbox1\)regularization

   Sebastian Buschjäger, and Katharina Morik

   Data Min. Knowl. Discov., Oct 2023

   PAPER Code Poster Slides Bibtex

   @article{Buschjaeger/Morik/2023,
     author = {Buschj{\"{a}}ger, Sebastian and Morik, Katharina},
     title = {Joint leaf-refinement and ensemble pruning through L\({}_{\mbox{1}}\) regularization},
     number = {3},
     pages = {1230--1261},
     volume = {37},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/journals/datamine/BuschjagerM23.bib},
     year = {2023},
     doi = {10.1007/s10618-023-00921-z},
     journal = {Data Min. Knowl. Discov.},
     timestamp = {Sat, 13 May 2023 01:07:00 +0200},
     url = {https://doi.org/10.1007/s10618-023-00921-z},
   }

2. arXiv

   Fast Inference of Tree Ensembles on ARM Devices

   Simon Koschel, Sebastian Buschjäger, Claudio Lucchese, and 1 more author

   arXiv, Oct 2023

   arXiv PAPER Bibtex

   @article{Koschel/etal/2023,
     author = {Koschel, Simon and Buschj{\"{a}}ger, Sebastian and Lucchese, Claudio and Morik, Katharina},
     title = {Fast Inference of Tree Ensembles on {ARM} Devices},
     volume = {abs/2305.08579},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/journals/corr/abs-2305-08579.bib},
     year = {2023},
     doi = {10.48550/arXiv.2305.08579},
     eprint = {2305.08579},
     eprinttype = {arXiv},
     journal = {arXiv},
     timestamp = {Wed, 17 May 2023 15:47:36 +0200},
     url = {https://doi.org/10.48550/arXiv.2305.08579},
   }

2022

1. PhD Thesis

   Ensemble learning with discrete classifiers on small devices

   Sebastian Buschjäger

   Technical University of Dortmund, Germany, Oct 2022

   PAPER Bibtex

   @phdthesis{Buschjaeger/2022,
     author = {Buschj{\"{a}}ger, Sebastian},
     title = {Ensemble learning with discrete classifiers on small devices},
     url = {http://hdl.handle.net/2003/41132},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/phd/dnb/Buschjager22.bib},
     school = {Technical University of Dortmund, Germany},
     journal = {Dissertation, TU Dortmund},
     timestamp = {Fri, 20 Jan 2023 21:27:23 +0100},
     urn = {urn:nbn:de:101:1-2022111802423335923878},
     year = {2022},
   }

2. AAAI

   Shrub Ensembles for Online Classification

   Sebastian Buschjäger, Sibylle Hess, and Katharina Morik

   In Proceedings of the Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-22), Oct 2022

   arXiv PAPER Code Poster Slides Bibtex

   @inproceedings{Buschjaeger/etal/2022,
     author = {Buschjäger, Sebastian and Hess, Sibylle and Morik, Katharina},
     booktitle = {Proceedings of the Thirty-Sixth {AAAI} Conference on Artificial Intelligence (AAAI-22)},
     title = {Shrub Ensembles for Online Classification},
     year = {2022},
     publisher = {{AAAI} Press},
   }

3. TCAS-I

   Reliable Binarized Neural Networks on Unreliable Beyond Von-Neumann Architecture

   Mikail Yayla, Simon Thomann, Sebastian Buschjäger, and 3 more authors

   IEEE Trans. Circuits Syst. I Regul. Pap., Oct 2022

   PAPER Bibtex

   @article{Yayla/etal/2022,
     author = {Yayla, Mikail and Thomann, Simon and Buschj{\"{a}}ger, Sebastian and Morik, Katharina and Chen, Jian{-}Jia and Amrouch, Hussam},
     journal = {{IEEE} Trans. Circuits Syst. {I} Regul. Pap.},
     title = {Reliable Binarized Neural Networks on Unreliable Beyond Von-Neumann Architecture},
     year = {2022},
     number = {6},
     pages = {2516--2528},
     volume = {69},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/journals/tcasI/YaylaTBMCA22.bib},
     doi = {10.1109/TCSI.2022.3156165},
     timestamp = {Mon, 28 Aug 2023 21:37:53 +0200},
     url = {https://doi.org/10.1109/TCSI.2022.3156165}
   }

4. Book Chapter

   Summary Extraction from Streams

   Sebastian Buschjäger, and Katharina Morik

   In Machine Learning under Resource Constraints - Volume 1: Fundamentals, Oct 2022

   PAPER Bibtex

   @incollection{Buschjaeger/Morik/2022,
     author = {Buschj{\"{a}}ger, Sebastian and Morik, Katharina},
     booktitle = {Machine Learning under Resource Constraints - Volume 1: Fundamentals},
     publisher = {De Gruyter},
     title = {Summary Extraction from Streams},
     year = {2022},
     editor = {Morik, Katharina and Marwedel, Peter},
     pages = {73--84},
     series = {De Gruyter {STEM}},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/books/degruyter/22/BuschjagerMM22-1.bib},
     doi = {10.1515/9783110785944-003},
     timestamp = {Fri, 28 Jul 2023 13:34:21 +0200},
     url = {https://doi.org/10.1515/9783110785944-003}
   }

5. Book Chapter

   Monitoring and Feature Extraction

   Wolfgang Rhode, Tim Ruhe, Maximilian Linhoff, and 3 more authors

   In Machine Learning under Resource Constraints - Volume 2: Discovery in Physics, Oct 2022

   PAPER Bibtex

   @incollection{Rhode/etal/2022,
     author = {Rhode, Wolfgang and Ruhe, Tim and Linhoff, Maximilian and Bu{\ss}, Jens and Nickel, Lukas and Buschj{\"{a}}ger, Sebastian},
     booktitle = {Machine Learning under Resource Constraints - Volume 2: Discovery in Physics},
     publisher = {De Gruyter},
     title = {Monitoring and Feature Extraction},
     year = {2022},
     editor = {Morik, Katharina and Rhode, Wolfgang},
     pages = {163--191},
     series = {De Gruyter {STEM}},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/books/degruyter/22/RhodeRLBNB22.bib},
     doi = {10.1515/9783110785968-007},
     timestamp = {Fri, 28 Jul 2023 11:18:35 +0200},
     url = {https://doi.org/10.1515/9783110785968-007}
   }

6. Book Chapter

   Deep Learning Applications

   Wolfgang Rhode, Mirco Hünnefeld, Bernhard Spaan, and 3 more authors

   In Machine Learning under Resource Constraints - Volume 2: Discovery in Physics, Oct 2022

   PAPER Bibtex

   @incollection{Rhode/etal/2022a,
     author = {Rhode, Wolfgang and H{\"{u}}nnefeld, Mirco and Spaan, Bernhard and Jevtic, Vukan and Pfahler, Lukas and Buschj{\"{a}}ger, Sebastian},
     booktitle = {Machine Learning under Resource Constraints - Volume 2: Discovery in Physics},
     publisher = {De Gruyter},
     title = {Deep Learning Applications},
     year = {2022},
     editor = {Morik, Katharina and Rhode, Wolfgang},
     pages = {245--278},
     series = {De Gruyter {STEM}},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/books/degruyter/22/RhodeHSJPB22.bib},
     doi = {10.1515/9783110785968-009},
     timestamp = {Fri, 28 Jul 2023 11:18:35 +0200},
     url = {https://doi.org/10.1515/9783110785968-009}
   }

7. Book Chapter

   Machine Learning Based on Emerging Memories

   Mikail Yayla, Sebastian Buschjäger, and Hussam Amrouch

   In Machine Learning under Resource Constraints - Volume 1: Fundamentals, Oct 2022

   PAPER Bibtex

   @incollection{Yayla/etal/2022a,
     author = {Yayla, Mikail and Buschj{\"{a}}ger, Sebastian and Amrouch, Hussam},
     booktitle = {Machine Learning under Resource Constraints - Volume 1: Fundamentals},
     publisher = {De Gruyter},
     title = {Machine Learning Based on Emerging Memories},
     year = {2022},
     editor = {Morik, Katharina and Marwedel, Peter},
     pages = {325--337},
     series = {De Gruyter {STEM}},
     bibsource = {dblp computer science bibliography, https://dblp.org},
     biburl = {https://dblp.org/rec/books/degruyter/22/KotthausMYBAC22-2.bib},
     doi = {10.1515/9783110785944-007},
     timestamp = {Fri, 28 Jul 2023 14:05:33 +0200},
     url = {https://doi.org/10.1515/9783110785944-007}
   }

2021

1. TECS

   Efficient Realization of Decision Trees for Real-Time Inference

   Kuan-Hsun Chen, ChiaHui Su, Christian Hakert, and 5 more authors

   ACM Transactions on Embedded Computing Systems, Dec 2021

   Abs PAPER Bibtex

   For timing-sensitive edge applications, the demand for efficient lightweight machine learning solutions has increased recently. Tree ensembles are among the state-of-the-art in many machine learning applications. While single decision trees are comparably small, an ensemble of trees can have a significant memory footprint leading to cache locality issues, which are crucial to performance in terms of execution time. In this work, we analyze memory-locality issues of the two most common realizations of decision trees, i.e. native and if-else trees. We highlight, that both realizations demand a more careful memory layout to improve caching behavior and maximize performance. We adopt a probabilistic model of decision tree inference to find the best memory layout for each tree at the application layer. Further, we present an efficient heuristic to take architecture-dependent information into account thereby optimizing the given ensemble for a target computer architecture. Our code-generation framework, which is freely available on an open-source repository, produces optimized code sessions while preserving the structure and accuracy of the trees. With several real-world data sets, we evaluate the elapsed time of various tree realizations on server hardware as well as embedded systems for Intel and ARM processors. Our optimized memory layout achieves a reduction in execution time up to 75 % execution for server-class systems, and up to 70 % for embedded systems, respectively.

   @article{Chen/etal/2021,
     author = {Chen, Kuan-Hsun and Su, ChiaHui and Hakert, Christian and Buschj\"{a}ger, Sebastian and Lee, Chao-Lin and Lee, Jenq-Kuen and Morik, Katharina and Chen, Jian-Jia},
     title = {Efficient Realization of Decision Trees for Real-Time Inference},
     doi = {10.1145/3508019},
     issn = {1539-9087},
     url = {https://doi.org/10.1145/3508019},
     address = {New York, NY, USA},
     journal = {{ACM} Transactions on Embedded Computing Systems},
     keywords = {Random Forest, Optimized Memory Layout, Real-Time Inference, Cache-Aware Optimization, Architecture-Aware Realization},
     month = dec,
     publisher = {Association for Computing Machinery},
     year = {2021},
   }

2. arXiv

   There is no Double-Descent in Random Forests

   Sebastian Buschjäger, and Katharina Morik

   arXiv, Dec 2021

   arXiv PAPER Bibtex

   @article{Buschjaeger/morik/2021c,
     author = {Buschj{\"{a}}ger, Sebastian and Morik, Katharina},
     title = {There is no Double-Descent in Random Forests},
     journal = {arXiv},
     volume = {abs/2111.04409},
     year = {2021},
     eprinttype = {arXiv},
     eprint = {2111.04409},
     timestamp = {Wed, 10 Nov 2021 16:07:30 +0100},
     biburl = {https://dblp.org/rec/journals/corr/abs-2111-04409.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org},
   }

3. arXiv

   Improving the Accuracy-Memory Trade-Off of Random Forests Via Leaf-Refinement

   Sebastian Buschjäger, and Katharina Morik

   arXiv, Dec 2021

   arXiv PAPER Bibtex

   @article{Buschjaeger/Morik/2021b,
     author = {Buschj{\"{a}}ger, Sebastian and Morik, Katharina},
     title = {Improving the Accuracy-Memory Trade-Off of Random Forests Via Leaf-Refinement},
     journal = {arXiv},
     volume = {abs/2110.10075},
     year = {2021},
     eprinttype = {arXiv},
     eprint = {2110.10075},
     timestamp = {Mon, 25 Oct 2021 20:07:12 +0200},
     biburl = {https://dblp.org/rec/journals/corr/abs-2110-10075.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org},
   }

4. ECML

   Very Fast Streaming Submodular Function Maximization

   Sebastian Buschjäger, Philipp-Jan Honysz, Lukas Pfahler, and 1 more author

   In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, Dec 2021

   PAPER Poster Slides Bibtex

   @inproceedings{Buschjaeger/etal/2021a,
     author = {Buschj\"{a}ger, Sebastian and Honysz, Philipp-Jan and Pfahler, Lukas and Morik, Katharina},
     title = {Very Fast Streaming Submodular Function Maximization},
     booktitle = {Joint European Conference on Machine Learning and Knowledge Discovery in Databases},
     publisher = {Springer},
     year = {2021},
   }

5. arXiv

   Providing Meaningful Data Summarizations Using Exemplar-based Clustering in Industry 4.0

   Philipp-Jan Honysz, Alexander Schulze-Struchtrup, Sebastian Buschjäger, and 1 more author

   arXiv, Dec 2021

   arXiv PAPER Bibtex

   @article{Honysz/etal/2021b,
     title = {Providing Meaningful Data Summarizations Using Exemplar-based Clustering in Industry 4.0},
     author = {Honysz, Philipp-Jan and Schulze-Struchtrup, Alexander and Buschjäger, Sebastian and Morik, Katharina},
     year = {2021},
     eprint = {2105.12026},
     journal = {arXiv}
   }

6. arXiv

   GPU-Accelerated Optimizer-Aware Evaluation of Submodular Exemplar Clustering

   Philipp-Jan Honysz, Sebastian Buschjäger, and Katharina Morik

   arXiv, Dec 2021

   Abs arXiv PAPER Code Bibtex

   The optimization of submodular functions constitutes a viable way to perform clustering. Strong approximation guarantees and feasible optimization w.r.t. streaming data make this clustering approach favorable. Technically, submodular functions map subsets of data to real values, which indicate how "representative" a specific subset is. Optimal sets might then be used to partition the data space and to infer clusters. Exemplar-based clustering is one of the possible submodular functions, but suffers from high computational complexity. However, for practical applications, the particular real-time or wall-clock run-time is decisive. In this work, we present a novel way to evaluate this particular function on GPUs, which keeps the necessities of optimizers in mind and reduces wall-clock run-time. To discuss our GPU algorithm, we investigated both the impact of different run-time critical problem properties, like data dimensionality and the number of data points in a subset, and the influence of required floating-point precision. In reproducible experiments, our GPU algorithm was able to achieve competitive speedups of up to 72x depending on whether multi-threaded computation on CPUs was used for comparison and the type of floating-point precision required. Half-precision GPU computation led to large speedups of up to 452x compared to single-precision, single-thread CPU computations.

   @article{Honysz/etal/2021a,
     title = {GPU-Accelerated Optimizer-Aware Evaluation of Submodular Exemplar Clustering},
     author = {Honysz, Philipp-Jan and Buschjäger, Sebastian and Morik, Katharina},
     year = {2021},
     journal = {arXiv}
   }

7. arxiv

   Bit Error Tolerance Metrics for Binarized Neural Networks

   Sebastian Buschjäger, Jian-Jia Chen, Kuan-Hsun Chen, and 5 more authors

   arXiv, Dec 2021

   arXiv PAPER Bibtex

   @article{Buschjaeger/etal/2021c,
     author = {Buschj{\"{a}}ger, Sebastian and Chen, Jian{-}Jia and Chen, Kuan{-}Hsun and G{\"{u}}nzel, Mario and Morik, Katharina and Novkin, Rodion and Pfahler, Lukas and Yayla, Mikail},
     title = {Bit Error Tolerance Metrics for Binarized Neural Networks},
     journal = {arXiv},
     volume = {abs/2102.01344},
     year = {2021},
     eprinttype = {arXiv},
     eprint = {2102.01344},
     timestamp = {Sun, 02 Oct 2022 15:32:08 +0200},
     biburl = {https://dblp.org/rec/journals/corr/abs-2102-01344.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org},
   }

8. TC

   FeFET-based Binarized Neural Networks Under Temperature-dependent Bit Errors

   Mikail Yayla, Sebastian Buschjager, Aniket Gupta, and 5 more authors

   IEEE Transactions on Computers, Dec 2021

   PAPER Bibtex

   @article{Yayla/etal/2021,
     author = {Yayla, Mikail and Buschjager, Sebastian and Gupta, Aniket and Chen, Jian-Jia and Henkel, Jorg and Morik, Katharina and Chen, Kuan-Hsun and Amrouch, Hussam},
     journal = {IEEE Transactions on Computers},
     title = {FeFET-based Binarized Neural Networks Under Temperature-dependent Bit Errors},
     year = {2021},
     pages = {1-1},
     doi = {10.1109/TC.2021.3104736},
     url = {https://ieeexplore.ieee.org/document/9513530}
   }

9. DATE

   Margin-Maximization in Binarized Neural Networks for Optimizing Bit Error Tolerance

   Sebastian Buschjäger, Jian-Jia Chen, Kuan-Hsun Chen, and 6 more authors

   In Design, Automation & Test in Europe Conference & Exhibition, DATE 2021, Grenoble, France, February 1-5, 2021, Dec 2021

   PAPER Bibtex

   @inproceedings{Buschjaeger/etal/2021b,
     author = {Buschj{\"{a}}ger, Sebastian and Chen, Jian{-}Jia and Chen, Kuan{-}Hsun and G{\"{u}}nzel, Mario and Hakert, Christian and Morik, Katharina and Novkin, Rodion and Pfahler, Lukas and Yayla, Mikail},
     title = {Margin-Maximization in Binarized Neural Networks for Optimizing Bit
                       Error Tolerance},
     booktitle = {Design, Automation {\&} Test in Europe Conference {\&} Exhibition,
                       {DATE} 2021, Grenoble, France, February 1-5, 2021},
     pages = {673--678},
     publisher = {{IEEE}},
     year = {2021},
     url = {https://doi.org/10.23919/DATE51398.2021.9473918},
     doi = {10.23919/DATE51398.2021.9473918},
     timestamp = {Sat, 30 Sep 2023 09:38:39 +0200},
     biburl = {https://dblp.org/rec/conf/date/BuschjagerCCGHM21.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org},
   }

2020

1. JDSA

   Randomized outlier detection with trees

   Sebastian Buschjäger, Philipp-jan Honysz, and Katharina Morik

   International Journal of Data Science and Analytics, Dec 2020

   PAPER Bibtex

   @article{Buschjager/Honysz/2020c,
     author = {Buschj{\"{a}}ger, Sebastian and Honysz, Philipp-jan and Morik, Katharina},
     doi = {10.1007/s41060-020-00238-w},
     isbn = {4106002000},
     issn = {2364-4168},
     journal = {International Journal of Data Science and Analytics},
     keywords = {Density estimation,Ensemble,Isolation forest,Outlier detection,Tree,density estimation,ensemble,isolation forest,outlier detection,tree},
     publisher = {Springer International Publishing},
     title = {{Randomized outlier detection with trees}},
     url = {https://doi.org/10.1007/s41060-020-00238-w},
     year = {2020},
     video = {https://www.youtube.com/watch?v=MOH6n8wiF7E},
   }

2. DSAA

   Generalized Isolation Forest: Some Theory and More Applications – Extended Abstract

   Sebastian Buschjäger, Philipp-Jan Honysz, and Katharina Morik

   In Proceedings 2020 IEEE 7th International Conference on Data Science and Advanced Analytics (DSAA 2020), Dec 2020

   Abs PAPER Bibtex

   Isolation Forest is a popular outlier detection algorithm that isolates outlier observations from regular observations by building multiple random decision trees. Multiple extensions enhance the original Isolation Forest algorithm including the Extended Isolation Forest which allows for non-rectangular splits and the SCiForest which improves the fitting of individual trees. All these approaches rate the outlierness of an observation by its average path-length. However, we find a lack of theoretical explanation on why these isolation-based algorithms offer such good practical performance. In this paper, we present a theoretical framework that describes the effectiveness of isolation-based approaches from a distributional viewpoint. We show that these algorithms fit a mixture of distributions, where the average path length of an observation can be viewed as a (somewhat crude) approximation of the mixture coefficient. Using this framework, we derive the Generalized Isolation Forest (GIF) which also trains random trees, but combining them moves beyond using the average path-length. In an extensive evaluation of over \350,000 experiments, we show that GIF outperforms the other methods on a variety of datasets while having comparable runtime.

   @inproceedings{Buschjaeger/Honysz/2020a,
     author = {Buschj\"{a}ger, Sebastian and Honysz, Philipp-Jan and Morik, Katharina},
     title = {Generalized Isolation Forest: Some Theory and More Applications -- Extended Abstract},
     booktitle = {Proceedings 2020 IEEE 7th International Conference on Data Science and Advanced Analytics (DSAA 2020)},
     organization = {IEEE},
     year = {2020},
     video = {https://www.youtube.com/watch?v=MOH6n8wiF7E},
   }

3. arXiv

   Very Fast Streaming Submodular Function Maximization

   Sebastian Buschjäger, Philipp-Jan Honysz, and Katharina Morik

   arXiv, Dec 2020

   Abs arXiv PAPER Code Bibtex

   Data summarization has become a valuable tool in understanding even terabytes of data. Due to their compelling theoretical properties, submodular functions have been in the focus of summarization algorithms. These algorithms offer worst-case approximations guarantees to the expense of higher computation and memory requirements. However, many practical applications do not fall under this worst-case, but are usually much more well-behaved. In this paper, we propose a new submodular function maximization algorithm called ThreeSieves, which ignores the worst-case, but delivers a good solution in high probability. It selects the most informative items from a data-stream on the fly and maintains a provable performance on a fixed memory budget. In an extensive evaluation of more than 7000 experiments, we show that our algorithm outperforms current state-of-the-art algorithms and, at the same time, uses fewer resources. Last, we highlight a real-world use-case of our algorithm for data summarization in gamma-ray astronomy.

   @article{Buschjaeger/Honysz/2020b,
     author = {Buschj\"{a}ger, Sebastian and Honysz, Philipp-Jan and Morik, Katharina},
     title = {Very Fast Streaming Submodular Function Maximization},
     year = {2020},
     journal = {arXiv}
   }

4. arXiv

   Generalized Negative Correlation Learning for Deep Ensembling

   Sebastian Buschjäger, Lukas Pfahler, and Katharina Morik

   arXiv, Dec 2020

   arXiv PAPER Bibtex

   @article{Buschjaeger/etal/2020c,
     author = {Buschj{\"{a}}ger, Sebastian and Pfahler, Lukas and Morik, Katharina},
     title = {Generalized Negative Correlation Learning for Deep Ensembling},
     journal = {arXiv},
     volume = {abs/2011.02952},
     year = {2020},
     eprinttype = {arXiv},
     eprint = {2011.02952},
     timestamp = {Mon, 09 Nov 2020 10:30:14 +0100},
     biburl = {https://dblp.org/rec/journals/corr/abs-2011-02952.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org},
   }

5. arXiv

   Towards Explainable Bit Error Tolerance of Resistive RAM-Based Binarized Neural Networks

   Sebastian Buschjäger, Jian-Jia Chen, Kuan-Hsun Chen, and 6 more authors

   arXiv, Dec 2020

   arXiv PAPER Bibtex

   @article{Buschjaeger/etal/2020b,
     author = {Buschj{\"{a}}ger, Sebastian and Chen, Jian{-}Jia and Chen, Kuan{-}Hsun and G{\"{u}}nzel, Mario and Hakert, Christian and Morik, Katharina and Novkin, Rodion and Pfahler, Lukas and Yayla, Mikail},
     title = {Towards Explainable Bit Error Tolerance of Resistive RAM-Based Binarized
                       Neural Networks},
     journal = {arXiv},
     volume = {abs/2002.00909},
     year = {2020},
     eprinttype = {arXiv},
     eprint = {2002.00909},
     timestamp = {Sun, 02 Oct 2022 15:31:59 +0200},
     biburl = {https://dblp.org/rec/journals/corr/abs-2002-00909.bib},
     bibsource = {dblp computer science bibliography, https://dblp.org}
   }

6. ECML

   On-Site Gamma-Hadron Separation with Deep Learning on FPGAs

   Sebastian Buschjäger, Lukas Pfahler, Jens Buss, and 2 more authors

   In Joint European Conference on Machine Learning and Knowledge Discovery in Databases, Dec 2020

   PAPER Slides Bibtex

   @inproceedings{Buschjaeger/etal/2020a,
     author = {Buschj\"{a}ger, Sebastian and Pfahler, Lukas and Buss, Jens and Morik, Katharina and Rhode, Wolfgang},
     title = {On-Site Gamma-Hadron Separation with Deep Learning on FPGAs},
     booktitle = {Joint European Conference on Machine Learning and Knowledge Discovery in Databases},
     publisher = {Springer},
     year = {2020},
   }

2019

1. MLCAD

   Stack Usage Analysis for Efficient Wear Leveling in Non-Volatile Main Memory Systems

   Christian Hakert, Mikail Yayla, Kuan-Hsun Chen, and 8 more authors

   In 1st ACM/IEEE Workshop on Machine Learning for CAD (MLCAD), Dec 2019

   PAPER Bibtex

   @inproceedings{Hakert/Yayla/2019a,
     author = {Hakert, Christian and Yayla, Mikail and Chen, Kuan-Hsun and Br\"uggen, Georg von der and Chen, Jian-Jia and Buschj\"ager, Sebastian and Morik, Katharina and Genssler, Paul R. and Bauer, Lars and Amrouch, Hussam and Henkel, J\"org},
     title = {Stack Usage Analysis for Efficient Wear Leveling in Non-Volatile Main Memory Systems},
     booktitle = {1st ACM/IEEE Workshop on Machine Learning for CAD (MLCAD)},
     year = {2019},
   }

2. ICPRAM

   Gaussian Model Trees for Traffic Imputation

   Sebastian Buschjäger, Thomas Liebig, and Katharina Morik

   In Proceedings of the International Conference on Pattern Recognition Applications and Methods (ICPRAM), Dec 2019

   Abs PAPER Slides Bibtex

   Traffic congestion is one of the most pressing issues for smart cities. Information on traffic flow can be used to reduce congestion by predicting vehicle counts at unmonitored locations so that counter-measures can be applied before congestion appears. To do so pricy sensors must be distributed sparsely in the city and at important roads in the city center to collect road and vehicle information throughout the city in real-time. Then, Machine Learning models can be applied to predict vehicle counts at unmonitored locations. To be fault-tolerant and increase coverage of the traffic predictions to the suburbs, rural regions, or even neighboring villages, these Machine Learning models should not operate at a central traffic control room but rather be distributed across the city. Gaussian Processes (GP) work well in the context of traffic count prediction, but cannot capitalize on the vast amount of data available in an entire city. Furthermore, Gaussian Processes are a global and centra lized model, which requires all measurements to be available at a central computation node. Product of Expert (PoE) models have been proposed as a scalable alternative to Gaussian Processes. A PoE model trains multiple, independent GPs on different subsets of the data and weight individual predictions based on each experts uncertainty. These methods work well, but they assume that experts are independent even though they may share data points. Furthermore, PoE models require exhaustive communication bandwidth between the individual experts to form the final prediction. In this paper we propose a hierarchical Product of Expert model, which consist of multiple layers of small, independent and local GP experts. We view Gaussian Process induction as regularized optimization procedure and utilize this view to derive an efficient algorithm which selects independent regions of the data. Then, we train local expert models on these regions, so that each expert is responsible for a given region. The resulting algorithm scales well for large amounts of data and outperforms flat PoE models in terms of communication cost, model size and predictive performance. Last, we discuss how to deploy these local expert models onto small devices

   @inproceedings{Buschjaeger/etal/2019a,
     author = {Buschj\"{a}ger, Sebastian and Liebig, Thomas and Morik, Katharina},
     title = {Gaussian Model Trees for Traffic Imputation},
     booktitle = {Proceedings of the International Conference on Pattern Recognition Applications and Methods (ICPRAM)},
     organization = {SciTePress},
     pages = {243 - 254},
     year = {2019},
     url = {https://www.scitepress.org/PublicationsDetail.aspx?ID=g+tVIY+KNts=\&t=1},
   }

2018

1. ICDM

   Realization of Random Forest for Real-Time Evaluation through Tree Framing

   Sebastian Buschjaeger, Kuan-Hsun Chen, Jian-Jia Chen, and 1 more author

   In The IEEE International Conference on Data Mining series (ICDM), Nov 2018

   Abs PAPER Poster Slides Bibtex

   The optimization of learning has always been of particular concern for big data analytics. However, the ongoing integration of machine learning models into everyday life also demand the evaluation to be extremely fast and in real-time. Moreover, in the Internet of Things, the computing facilities that run the learned model are restricted. Hence, the implementation of the model application must take the characteristics of the executing platform into account Although there exist some heuristics that optimize the code, principled approaches for fast execution of learned models are rare. In this paper, we introduce a method that optimizes the execution of Decision Trees (DT). Decision Trees form the basis of many ensemble methods, such as Random Forests (RF) or Extremely Randomized Trees (ET). For these methods to work best, trees should be as large as possible. This challenges the data and the instruction cache of modern CPUs and thus demand a more careful memory layout. Based on a probabilistic view of decision tree execution, we optimize the two most common implementation schemes of decision trees. We discuss the advantages and disadvantages of both implementations and present a theoretically well-founded memory layout which maximizes locality during execution in both cases. The method is applied to three computer architectures, namely ARM (RISC), PPC (Extended RISC) and Intel (CISC) and is automatically adopted to the specific architecture by a code generator. We perform over 1800 experiments on several real-world data sets and report an average speed-up of 2 to 4 across all three architectures by using the proposed memory layout. Moreover, we find that our implementation outperforms sklearn, which was used to train the models by a factor of 1500.

   @inproceedings{Buschjaeger/2018a,
     author = {Buschjaeger, Sebastian and Chen, Kuan-Hsun and Chen, Jian-Jia and Morik, Katharina},
     title = {Realization of Random Forest for Real-Time  Evaluation through Tree Framing},
     booktitle = {The IEEE International Conference on Data Mining series (ICDM)},
     month = nov,
     year = {2018},
     url = {https://ieeexplore.ieee.org/document/8594826},
   }

2. TCAS-I

   Decision Tree and Random Forest Implementations for Fast Filtering of Sensor Data

   Sebastian Buschjäger, and Katharina Morik

   IEEE Transactions on Circuits and Systems I: Regular Papers, Jan 2018

   PAPER Bibtex

   @article{Buschjaeger/Morik/2017b,
     author = {Buschj\"{a}ger, Sebastian and Morik, Katharina},
     title = {Decision Tree and Random Forest Implementations for Fast Filtering of Sensor Data},
     journal = {IEEE Transactions on Circuits and Systems I: Regular Papers},
     month = jan,
     number = {1},
     pages = {209--222},
     volume = {65-I},
     year = {2018},
     url = {https://doi.org/10.1109/TCSI.2017.2710627},
   }

2017

1. KI

   Big Data Science

   Katharina Morik, Christian Bockermann, and Sebastian Buschjäger

   German journal on Artificial Intelligence, Dec 2017

   Abs PAPER Bibtex

   In ever more disciplines, science is driven by data, which leads to data analytics becoming a primary skill for researchers. This includes the complete process from data acquisition at sensors, over pre-processing and feature extraction to the use and application of machine learning. Sensors here often produce a plethora of data that needs to be dealt with in near-realtime, which requires a combined effort of implementations at the hardware level to high-level design of data flows. In this paper we outline two use-cases of this wide span of data analysis for science in a real-world example in astroparticle physics. We outline a high-level design approach which is capable of defining the complete data flow from sensor hardware to final analysis.

   @article{Morik/etal/2017a,
     author = {Morik, Katharina and Bockermann, Christian and Buschj\"{a}ger, Sebastian},
     title = {Big Data Science},
     journal = {German journal on Artificial Intelligence},
     month = dec,
     number = {1},
     pages = {27--36},
     volume = {32},
     year = {2017},
     url = {https://doi.org/10.1007/s13218-017-0522-8},
   }

2. IoTStreaming

   Summary Extraction on Data Streams in Embedded Systems

   Sebastian Buschjäger, Katharina Morik, and Maik Schmidt

   In Proceedings of the ECML Workshop on IoT Large Scale Learning From Data Streams, Dec 2017

   PAPER Slides Bibtex

   @inproceedings{Buschjaeger/Morik/2017a,
     author = {Buschj\"{a}ger, Sebastian and Morik, Katharina and Schmidt, Maik},
     title = {Summary Extraction on Data Streams in Embedded Systems},
     booktitle = {Proceedings of the ECML Workshop on IoT Large Scale Learning From Data Streams},
     publisher = {ceur-ws.org},
     year = {2017},
   }

2016

1. MA

   Online Gauß-Prozesse zur Regression auf FPGAs

   Sebastian Buschjäger

   Dec 2016

   PAPER Bibtex

   @masterthesis{Buschjaeger/2016a,
     author = {Buschj\"{a}ger, Sebastian},
     title = {Online Gau{\ss}-Prozesse zur Regression auf FPGAs},
     school = {TU Dortmund},
     journal = {Masterthesis, TU Dortmund},
     year = {2016},
   }

2015

1. Discovering Subtle Word Relation in Large German Corpora

   Sebastian Buschjäger, Lukas Pfahler, and Katharina Morik

   In Proceedings of the 3rd Workshop on the Challenges in the Management of Large Corpora, Dec 2015

   PAPER Bibtex

   @inproceedings{Buschjaeger/etal/2015a,
     author = {Buschj\"{a}ger, Sebastian and Pfahler, Lukas and Morik, Katharina},
     title = {Discovering Subtle Word Relation in Large German Corpora},
     booktitle = {Proceedings of the 3rd Workshop on the Challenges in the Management of Large Corpora},
     year = {2015},
   }

2. Untersuchungen zur Analyse von deutschsprachigen Textdaten

   Katharina Morik, Alexander Jung, Jan Weckwerth, and 4 more authors

   Dec 2015

   PAPER Bibtex

   @techreport{Morik/etal/2015a,
     author = {Morik, Katharina and Jung, Alexander and Weckwerth, Jan and R{\"o}tner, Stefan and Hess, Sibylle and Buschj\"{a}ger, Sebastian and Pfahler, Lukas},
     title = {Untersuchungen zur Analyse von deutschsprachigen Textdaten},
     institution = {Technische Universit\"{a}t Dortmund},
     month = dec,
     number = {2},
     year = {2015},
   }