TY - CONF TI - Intrinsic User-Centric Interpretability through Global Mixture of Experts AU - Swamy, Vinitra AU - Montariol, Syrielle AU - Blackwell, Julian AU - Frej, Jibril AU - Jaggi, Martin AU - Käser, Tanja T2 - The Thirteenth International Conference on Learning Representations AB - In human-centric settings like education or healthcare, model accuracy and model explainability are key factors for user adoption. Towards these two goals, intrinsically interpretable deep learning models have gained popularity, focusing on accurate predictions alongside faithful explanations. However, there exists a gap in the human-centeredness of these approaches, which often produce nuanced and complex explanations that are not easily actionable for downstream users. We present InterpretCC (interpretable conditional computation), a family of intrinsically interpretable neural networks at a unique point in the design space that optimizes for ease of human understanding and explanation faithfulness, while maintaining comparable performance to state-of-the-art models. InterpretCC achieves this through adaptive sparse activation of features before prediction, allowing the model to use a different, minimal set of features for each instance. We extend this idea into an interpretable, global mixture-of-experts (MoE) model that allows users to specify topics of interest, discretely separates the feature space for each data point into topical subnetworks, and adaptively and sparsely activates these topical subnetworks for prediction. We apply InterpretCC for text, time series and tabular data across several real-world datasets, demonstrating comparable performance with non-interpretable baselines and outperforming intrinsically interpretable baselines. Through a user study involving 56 teachers, InterpretCC explanations are found to have higher actionability and usefulness over other intrinsically interpretable approaches. DA - 2024/10/04/ PY - 2024 DP - openreview.net LA - en UR - https://openreview.net/forum?id=wDcunIOAOk Y2 - 2025/05/19/09:34:15 ER - TY - JOUR TI - Vacuum-sealed silicon photonic MEMS tunable ring resonator with an independent control over coupling and phase AU - Edinger, Pierre AU - Jo, Gaehun AU - Van Nguyen, Chris Phong AU - Takabayashi, Alain Yuji AU - Errando-Herranz, Carlos AU - Antony, Cleitus AU - Talli, Giuseppe AU - Verheyen, Peter AU - Khan, Umar AU - Bleiker, Simon J. AU - Bogaerts, Wim AU - Quack, Niels AU - Niklaus, Frank AU - Gylfason, Kristinn B. T2 - Optics Express AB - Ring resonators are a vital element for filters, optical delay lines, or sensors in silicon photonics. However, reconfigurable ring resonators with low-power consumption are not available in foundries today. We demonstrate an add-drop ring resonator with the independent tuning of round-trip phase and coupling using low-power microelectromechanical (MEMS) actuation. At a wavelength of 1540 nm and for a maximum voltage of 40 V, the phase shifters provide a resonance wavelength tuning of 0.15 nm, while the tunable couplers can tune the optical resonance extinction ratio at the through port from 0 to 30 dB. The optical resonance displays a passive quality factor of 29 000, which can be increased to almost 50 000 with actuation. The MEMS rings are individually vacuum-sealed on wafer scale, enabling reliable and long-term protection from the environment. We cycled the mechanical actuators for more than 4 × 10 9 cycles at 100 kHz, and did not observe degradation in their response curves. On mechanical resonance, we demonstrate a modulation increase of up to 15 dB, with a voltage bias of 4 V and a peak drive amplitude as low as 20 mV. DA - 2023/02/13/ PY - 2023 DO - 10.1364/OE.480219 DP - DOI.org (Crossref) VL - 31 IS - 4 SP - 6540 J2 - Opt. Express LA - en SN - 1094-4087 UR - https://opg.optica.org/abstract.cfm?URI=oe-31-4-6540 AN - http://hdl.handle.net/1854/LU-01GSFC65JQ2F45VQ2SAV9N9F7W DB - Ghent University Library Y2 - 2025/03/31/14:10:48 ER - TY - JOUR TI - pH Quantification in Human Dermal Interstitial Fluid Using Ultra-Thin SOI Silicon Nanowire ISFETs and a High-Sensitivity Constant-Current Approach AU - Sprunger, Yann AU - Capua, Luca AU - Ernst, Thomas AU - Barraud, Sylvain AU - Locca, Didier AU - Ionescu, Adrian AU - Saeidi, Ali T2 - Biosensors AB - In this paper, we propose a novel approach to utilize silicon nanowires as high-sensitivity pH sensors. Our approach works based on fixing the current bias of silicon nanowires Ion Sensitive Field Effect Transistors (ISFETs) and monitor the resulting drain voltage as the sensing signal. By fine tuning the injected current levels, we can optimize the sensing conditions according to different sensor requirements. This method proves to be highly suitable for real-time and continuous measurements of biomarkers in human biofluids. To validate our approach, we conducted experiments, with real human sera samples to simulate the composition of human interstitial fluid (ISF), using both the conventional top-gate approach and the optimized constant current method. We successfully demonstrated pH sensing within the physiopathological range of 6.5 to 8, achieving an exceptional level of accuracy in this complex matrix. Specifically, we obtained a maximum error as low as 0.92% (equivalent to 0.07 pH unit) using the constant-current method at the optimal current levels (1.71% for top-gate). Moreover, by utilizing different pools of human sera with varying total protein content, we demonstrated that the protein content among patients does not impact the sensors’ performance in pH sensing. Furthermore, we tested real-human ISF samples collected from volunteers. The obtained accuracy in this scenario was also outstanding, with an error as low as 0.015 pH unit using the constant-current method and 0.178 pH unit in traditional top-gate configuration. DA - 2023/09/27/ PY - 2023 DO - 10.3390/bios13100908 DP - DOI.org (Crossref) VL - 13 IS - 10 SP - 908 J2 - Biosensors LA - en SN - 2079-6374 UR - https://www.mdpi.com/2079-6374/13/10/908 AN - https://pmc.ncbi.nlm.nih.gov/articles/PMC10605508/ DB - PMC Y2 - 2024/05/15/12:02:41 ER - TY - CONF TI - Model-Based ISO 14971 Risk Management of EEG-Based Medical Devices AU - Yakymets, N. AU - Zanetti, R. AU - Ionescu, A. AU - Atienza, D. T2 - 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) C1 - Sydney, Australia C3 - 2023 45th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) DA - 2023/07/24/ PY - 2023 DO - 10.1109/EMBC40787.2023.10340131 DP - DOI.org (Crossref) SP - 1 EP - 7 PB - IEEE SN - 979-8-3503-2447-1 UR - https://ieeexplore.ieee.org/document/10340131/ AN - https://infoscience.epfl.ch/record/299748?v=pdf#files DB - Infoscience Y2 - 2024/05/15/12:06:21 ER - TY - JOUR TI - TimEHR: Image-based Time Series Generation for Electronic Health Records AU - Karami, Hojjat AU - Hartley, Mary-Anne AU - Atienza, David AU - Ionescu, Anisoara T2 - IEEE Journal of Biomedical and Health Informatics DA - 2025/// PY - 2025 DO - 10.1109/JBHI.2025.3577328 DP - DOI.org (Crossref) SP - 1 EP - 12 J2 - IEEE J. Biomed. Health Inform. SN - 2168-2194, 2168-2208 ST - TimEHR UR - https://ieeexplore.ieee.org/document/11027528/ AN - https://arxiv.org/abs/2402.06318 DB - arXiv Y2 - 2025/08/30/20:33:59 ER - TY - CONF TI - Ultra-High Sensitivity Silicon Nanowire Array Biosensor Based on a Constant-Current Method for Continuous Real-Time pH and Protein Monitoring in Interstitial Fluid AU - Sprunger, Y. AU - Capua, L. AU - Ernst, T. AU - Barraud, S. AU - Ionescu, A.M. AU - Saeidi, A. T2 - ESSCIRC 2023- IEEE 49th European Solid State Circuits Conference (ESSCIRC) C1 - Lisbon, Portugal C3 - ESSCIRC 2023- IEEE 49th European Solid State Circuits Conference (ESSCIRC) DA - 2023/09/11/ PY - 2023 DO - 10.1109/ESSCIRC59616.2023.10268731 DP - DOI.org (Crossref) SP - 153 EP - 156 PB - IEEE SN - 979-8-3503-0420-6 UR - https://ieeexplore.ieee.org/document/10268731/ AN - https://infoscience.epfl.ch/handle/20.500.14299/203930 DB - Infoscience Y2 - 2023/12/07/20:19:29 ER - TY - JOUR TI - Use of a Silicon Microneedle Chip-Based Device for the Extraction and Subsequent Analysis of Dermal Interstitial Fluid in Heart Failure Patients AU - Renlund, Markus AU - Kopp Fernandes, Laurenz AU - Rangsten, Pelle AU - Hillmering, Mikael AU - Mosel, Sara AU - Issa, Ziad AU - Falk, Volkmar AU - Meyer, Alexander AU - Schoenrath, Felix T2 - Diagnostics AB - Background/Objectives: Dermal interstitial fluid (dISF) is probably the most interesting biofluid for biomarker analysis as an alternative to blood, enabling higher patient comfort and closer or even continuous biomarker monitoring. The prerequisite for dISF-based analysis tools is having convenient access to dISF, as well as a better knowledge of the presence, concentration, and dynamics of biomarkers in dISF. Hollow microneedles represent one of the most promising platforms for access to pure dISF, enabling the mining of biomarker information. Methods and Results: Here, a microneedle-based method for dISF sampling is presented, where a combination of hollow microneedles and sub-pressure is used to optimize both penetration depth in skin and dermal interstitial fluid sampling volumes, and the design of an open, prospective, exploratory, and interventional study to examine the detectability of inflammatory and cardiocirculatory biomarkers in the dISF of heart failure patients, the relationship between dISF-derived and blood-derived biomarker levels, and their kinetics during a cardiopulmonary exercise test (CPET) is introduced. Conclusions: The dISF sampling method and study presented here will foster research on biomarkers in dISF in general and in heart failure patients in particular. The study is part of the European project DIGIPREDICT—Digital Edge AI-deployed DIGItal Twins for PREDICTing disease progression and the need for early intervention in infectious and cardiovascular diseases beyond COVID-19. DA - 2025/04/13/ PY - 2025 DO - 10.3390/diagnostics15080989 DP - DOI.org (Crossref) VL - 15 IS - 8 SP - 989 J2 - Diagnostics LA - en SN - 2075-4418 UR - https://www.mdpi.com/2075-4418/15/8/989 AN - https://www.scilit.com/publications/e8521adb75656e449d6f76ec7c79ac7c DB - scilit Y2 - 2025/04/23/09:20:00 ER - TY - CONF TI - Integrated Silicon-On-Insulator Based Mesh Membrane for Continuous Monitoring in Organs-on-a-chip AU - Zakharova, Mariia AU - Cóndor, Mar AU - Shaikh, Sohail F. AU - Delahanty, Aaron AU - Braeken, Dries AU - Van Der Meer, Andries D. AU - Segerink, Loes I. T2 - 2024 IEEE SENSORS C1 - Kobe, Japan C3 - 2024 IEEE SENSORS DA - 2024/10/20/ PY - 2024 DO - 10.1109/SENSORS60989.2024.10784460 DP - DOI.org (Crossref) SP - 1 EP - 4 PB - IEEE SN - 979-8-3503-6351-7 UR - https://ieeexplore.ieee.org/document/10784460/ AN - https://research.utwente.nl/en/publications/integrated-silicon-on-insulator-based-mesh-membrane-for-continuou DB - utwente Y2 - 2025/04/15/16:03:01 ER - TY - CONF TI - Point-process-based Representation Learning for Electronic Health Records AU - Karami, Hojjat AU - Ionescu, Anisoara AU - Atienza, David T2 - 2023 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI) C1 - Pittsburgh, PA, USA C3 - 2023 IEEE EMBS International Conference on Biomedical and Health Informatics (BHI) DA - 2023/10/15/ PY - 2023 DO - 10.1109/BHI58575.2023.10313499 DP - DOI.org (Crossref) SP - 1 EP - 4 PB - IEEE SN - 979-8-3503-1050-4 UR - https://ieeexplore.ieee.org/document/10313499/ AN - https://infoscience.epfl.ch/record/301541 DB - Infoscience Y2 - 2023/12/07/20:13:42 ER - TY - CONF TI - How to Count Coughs: An Event-Based Framework for Evaluating Automatic Cough Detection Algorithm Performance AU - Orlandic, Lara AU - Dan, Jonathan AU - Thevenot, Jérôme AU - Teijeiro, Tomas AU - Sauty, Alain AU - Atienza, David T2 - 2024 IEEE 20th International Conference on Body Sensor Networks (BSN) C1 - Chicago, IL, USA C3 - 2024 IEEE 20th International Conference on Body Sensor Networks (BSN) DA - 2024/10/15/ PY - 2024 DO - 10.1109/BSN63547.2024.10780617 DP - DOI.org (Crossref) SP - 1 EP - 4 PB - IEEE SN - 979-8-3315-3014-3 ST - How to Count Coughs UR - https://ieeexplore.ieee.org/document/10780617/ AN - https://arxiv.org/abs/2406.01529 DB - arXiv Y2 - 2025/01/17/14:49:32 ER - TY - JOUR TI - TEE4EHR: Transformer event encoder for better representation learning in electronic health records AU - Karami, Hojjat AU - Atienza, David AU - Ionescu, Anisoara T2 - Artificial Intelligence in Medicine DA - 2024/08// PY - 2024 DO - 10.1016/j.artmed.2024.102903 DP - DOI.org (Crossref) VL - 154 SP - 102903 J2 - Artificial Intelligence in Medicine LA - en SN - 09333657 ST - TEE4EHR UR - https://linkinghub.elsevier.com/retrieve/pii/S0933365724001453 AN - https://arxiv.org/abs/2402.06367 DB - arXiv Y2 - 2024/11/26/15:01:14 ER - TY - CONF TI - Tunable Dual Mode Carbon Nanotube Strain Gauge AU - Vollmann, Morten AU - Roman, Cosmin AU - Hierold, Christofer T2 - 2024 IEEE 37th International Conference on Micro Electro Mechanical Systems (MEMS) C1 - Austin, TX, USA C3 - 2024 IEEE 37th International Conference on Micro Electro Mechanical Systems (MEMS) DA - 2024/01/21/ PY - 2024 DO - 10.1109/MEMS58180.2024.10439427 DP - DOI.org (Crossref) SP - 931 EP - 934 PB - IEEE SN - 979-8-3503-5792-9 UR - https://ieeexplore.ieee.org/document/10439427/ Y2 - 2024/11/26/14:54:50 ER - TY - JOUR TI - A Multichannel Electrochemical Sensor Interface IC for Bioreactor Monitoring AU - Lin, Qiuyang AU - Sijbers, Wim AU - Avdikou, Christina AU - Gomez, Didac AU - Biswas, Dwaipayan AU - Tacca, Bernardo AU - Van Helleputte, Nick T2 - IEEE Transactions on Biomedical Circuits and Systems DA - 2023/// PY - 2023 DO - 10.1109/TBCAS.2023.3315480 DP - DOI.org (Crossref) SP - 1 EP - 9 J2 - IEEE Trans. Biomed. Circuits Syst. SN - 1932-4545, 1940-9990 UR - https://ieeexplore.ieee.org/document/10251576/ AN - https://zenodo.org/records/8389210 DB - Zenodo Y2 - 2023/09/28/11:19:35 ER - TY - CONF TI - A Multimodal Dataset for Automatic Edge-AI Cough Detection AU - Orlandic, Lara AU - Thevenot, Jérôme AU - Teijeiro, Tomas AU - Atienza, David AB - Counting the number of times a patient coughs per day is an essential biomarker in determining treatment efficacy for novel antitussive therapies and personalizing patient care. There is a need for wearable devices that employ multimodal sensors to perform accurate, privacy-preserving, automatic cough counting algorithms directly on the device in an edge-AI fashion. To advance this research field, we contribute the first publicly accessible cough counting dataset of multimodal biosignals. The database contains nearly 4 hours of biosignal data, with both acoustic and kinematic modalities, covering 4,300 annotated cough events. Furthermore, several non-cough sounds (i.e. breathing, laughing, and throat clearing), background noises (i.e. music, traffic, bystander coughing) and motion scenarios (i.e. sitting, walking) mimicking daily life activities are also present, which the research community can use to accelerate ML algorithm development. For detailed information about using this dataset to train edge-AI models and example code, please refer to our public Git repository: https://github.com/esl-epfl/edge-ai-cough-count/ DA - 2023/01/23/ PY - 2023 DO - 10.5281/ZENODO.7562332 DP - DOI.org (Datacite) PB - IEEE UR - https://zenodo.org/record/7562332 AN - https://zenodo.org/record/7562332 DB - Zenodo Y2 - 2023/09/28/10:45:07 KW - automatic cough detection KW - edge-AI KW - multimodal biosignals ER - TY - JOUR TI - A semi-supervised algorithm for improving the consistency of crowdsourced datasets: The COVID-19 case study on respiratory disorder classification AU - Orlandic, Lara AU - Teijeiro, Tomas AU - Atienza, David T2 - Computer Methods and Programs in Biomedicine DA - 2023/11// PY - 2023 DO - 10.1016/j.cmpb.2023.107743 DP - DOI.org (Crossref) VL - 241 SP - 107743 J2 - Computer Methods and Programs in Biomedicine LA - en SN - 01692607 ST - A semi-supervised algorithm for improving the consistency of crowdsourced datasets UR - https://linkinghub.elsevier.com/retrieve/pii/S0169260723004091 AN - https://arxiv.org/pdf/2209.04360.pdf DB - arXiv Y2 - 2023/08/21/16:05:29 ER - TY - JOUR TI - Event-based sampled ECG morphology reconstruction through self-similarity AU - Zanoli, Silvio AU - Ansaloni, Giovanni AU - Teijeiro, Tomás AU - Atienza, David T2 - Computer Methods and Programs in Biomedicine DA - 2023/10// PY - 2023 DO - 10.1016/j.cmpb.2023.107712 DP - DOI.org (Crossref) VL - 240 SP - 107712 J2 - Computer Methods and Programs in Biomedicine LA - en SN - 01692607 UR - https://linkinghub.elsevier.com/retrieve/pii/S0169260723003784 AN - https://arxiv.org/abs/2207.01856 DB - arXiv Y2 - 2023/07/13/14:55:40 ER - TY - JOUR TI - An Error-Based Approximation Sensing Circuit for Event-Triggered Low-Power Wearable Sensors AU - Zanoli, Silvio AU - Ponzina, Flavio AU - Teijeiro, Tomás AU - Levisse, Alexandre AU - Atienza, David T2 - IEEE Journal on Emerging and Selected Topics in Circuits and Systems DA - 2023/06// PY - 2023 DO - 10.1109/JETCAS.2023.3269623 DP - DOI.org (Crossref) VL - 13 IS - 2 SP - 489 EP - 501 J2 - IEEE J. Emerg. Sel. Topics Circuits Syst. SN - 2156-3357, 2156-3365 UR - https://ieeexplore.ieee.org/document/10107413/ AN - https://arxiv.org/abs/2106.13545 DB - arXiv Y2 - 2023/07/06/15:22:37 ER - TY - JOUR TI - Prognostic models in COVID-19 infection that predict severity: a systematic review AU - Buttia, Chepkoech AU - Llanaj, Erand AU - Raeisi-Dehkordi, Hamidreza AU - Kastrati, Lum AU - Amiri, Mojgan AU - Meçani, Renald AU - Taneri, Petek Eylul AU - Ochoa, Sergio Alejandro Gómez AU - Raguindin, Peter Francis AU - Wehrli, Faina AU - Khatami, Farnaz AU - Espínola, Octavio Pano AU - Rojas, Lyda Z. AU - de Mortanges, Aurélie Pahud AU - Macharia-Nimietz, Eric Francis AU - Alijla, Fadi AU - Minder, Beatrice AU - Leichtle, Alexander B. AU - Lüthi, Nora AU - Ehrhard, Simone AU - Que, Yok-Ai AU - Fernandes, Laurenz Kopp AU - Hautz, Wolf AU - Muka, Taulant T2 - European Journal of Epidemiology AB - Abstract Current evidence on COVID-19 prognostic models is inconsistent and clinical applicability remains controversial. We performed a systematic review to summarize and critically appraise the available studies that have developed, assessed and/or validated prognostic models of COVID-19 predicting health outcomes. We searched six bibliographic databases to identify published articles that investigated univariable and multivariable prognostic models predicting adverse outcomes in adult COVID-19 patients, including intensive care unit (ICU) admission, intubation, high-flow nasal therapy (HFNT), extracorporeal membrane oxygenation (ECMO) and mortality. We identified and assessed 314 eligible articles from more than 40 countries, with 152 of these studies presenting mortality, 66 progression to severe or critical illness, 35 mortality and ICU admission combined, 17 ICU admission only, while the remaining 44 studies reported prediction models for mechanical ventilation (MV) or a combination of multiple outcomes. The sample size of included studies varied from 11 to 7,704,171 participants, with a mean age ranging from 18 to 93 years. There were 353 prognostic models investigated, with area under the curve (AUC) ranging from 0.44 to 0.99. A great proportion of studies (61.5%, 193 out of 314) performed internal or external validation or replication. In 312 (99.4%) studies, prognostic models were reported to be at high risk of bias due to uncertainties and challenges surrounding methodological rigor, sampling, handling of missing data, failure to deal with overfitting and heterogeneous definitions of COVID-19 and severity outcomes. While several clinical prognostic models for COVID-19 have been described in the literature, they are limited in generalizability and/or applicability due to deficiencies in addressing fundamental statistical and methodological concerns. Future large, multi-centric and well-designed prognostic prospective studies are needed to clarify remaining uncertainties. DA - 2023/02/25/ PY - 2023 DO - 10.1007/s10654-023-00973-x DP - DOI.org (Crossref) J2 - Eur J Epidemiol LA - en SN - 0393-2990, 1573-7284 ST - Prognostic models in COVID-19 infection that predict severity UR - https://link.springer.com/10.1007/s10654-023-00973-x AN - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9958330/ DB - PMC Y2 - 2023/03/09/13:56:34 ER - TY - CONF TI - Double-Gate Si Nanowire FET Sensor Arrays For Label-Free C-Reactive Protein detection enabled by antibodies fragments and pseudo-super-Nernstian back-gate operation AU - Capua, L. AU - Sprunger, Y. AU - Elettro, H. AU - Grammoustianou, A. AU - Midahuen, R. AU - Ernst, T. AU - Barraud, S. AU - Gill, R. AU - Ionescu, A.M. T2 - 2021 IEEE International Electron Devices Meeting (IEDM) C1 - San Francisco, CA, USA C3 - 2021 IEEE International Electron Devices Meeting (IEDM) DA - 2021/12/11/ PY - 2021 DO - 10.1109/IEDM19574.2021.9720670 DP - DOI.org (Crossref) SP - 16.2.1 EP - 16.2.4 PB - IEEE SN - 978-1-6654-2572-8 UR - https://ieeexplore.ieee.org/document/9720670/ AN - https://infoscience.epfl.ch/record/291466 DB - Infoscience Y2 - 2022/05/19/12:42:49 ER - TY - CONF TI - An Improved Analysis of Gradient Tracking for Decentralized Machine Learning AU - Koloskova, Anastasiia AU - Lin, Tao AU - Stich, Sebastian U. T2 - Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS 202) T3 - NeurIPS Proceedings C3 - Advances in Neural Information Processing Systems DA - 2021/12/06/ PY - 2021 DP - proceedings.neurips.cc VL - 34 LA - en PB - NIPS SN - 978-1-7138-4539-3 UR - https://proceedings.neurips.cc/paper/2021/hash/5f25fbe144e4a81a1b0080b6c1032778-Abstract.html AN - https://arxiv.org/abs/2202.03836 DB - arXiv Y2 - 2022/02/23/14:27:17 ER - TY - CONF TI - Consensus Control for Decentralized Deep Learning AU - Kong, Lingjing AU - Lin, Tao AU - Koloskova, Anastasia AU - Jaggi, Martin AU - Stich, Sebastian U T2 - 38th International Conference on Machine Learning (ICML) T3 - Proceedings of Machine Learning Research AB - Decentralized training of deep learning models enables on-device learning over networks, as well as efficient scaling to large compute clusters. Experiments in earlier works reveal that, even in a data-center setup, decentralized training often suffers from the degradation in the quality of the model: the training and test performance of models trained in a decentralized fashion is in general worse than that of models trained in a centralized fashion, and this performance drop is impacted by parameters such as network size, communication topology and data partitioning. C3 - Proceedings of the 38th International Conference on Machine Learning DA - 2021/// PY - 2021 VL - 139 SP - 5686 EP - 5696 LA - en PB - PMLR UR - https://proceedings.mlr.press/v139/kong21a.html AN - https://arxiv.org/abs/2102.04828 DB - arXiv ER - TY - CONF TI - RelaySum for Decentralized Deep Learning on Heterogeneous Data AU - Vogels, Thijs AU - He, Lie AU - Koloskova, Anastasiia AU - Karimireddy, Sai Praneeth AU - Lin, Tao AU - Stich, Sebastian U. AU - Jaggi, Martin T2 - Thirty-fifth Conference on Neural Information Processing Systems (NeurIPS 202) T3 - NeurIPS Proceedings C3 - Advances in Neural Information Processing Systems DA - 2021/12/06/ PY - 2021 DP - proceedings.neurips.cc VL - 34 LA - en PB - NIPS UR - https://proceedings.neurips.cc/paper/2021/hash/ebbdfea212e3a756a1fded7b35578525-Abstract.html AN - https://arxiv.org/abs/2110.04175 DB - arXiv Y2 - 2022/02/23/13:56:46 ER - TY - JOUR TI - Label-Free C-Reactive Protein Si Nanowire FET Sensor Arrays With Super-Nernstian Back-Gate Operation AU - Capua, Luca AU - Sprunger, Yann AU - Elettro, H. AU - Risch, F. AU - Grammoustianou, A. AU - Midahuen, R. AU - Ernst, T. AU - Barraud, S. AU - Gill, R. AU - Ionescu, A. M. T2 - IEEE Transactions on Electron Devices DA - 2022/// PY - 2022 DO - 10.1109/TED.2022.3144108 DP - DOI.org (Crossref) SP - 1 EP - 7 J2 - IEEE Trans. Electron Devices SN - 0018-9383, 1557-9646 UR - https://ieeexplore.ieee.org/document/9709497/ AN - http://infoscience.epfl.ch/record/292349 DB - Infoscience Y2 - 2022/02/16/14:56:09 KW - peer-reviewed ER - TY - CONF TI - Quasi-Global Momentum: Accelerating Decentralized Deep Learning on Heterogeneous Data AU - Lin, Tao AU - Karimireddy, Sai Praneeth AU - Stich, Sebastian U AU - Jaggi, Martin T2 - 38th International Conference on Machine Learning (ICML) T3 - Proceedings of Machine Learning Research AB - Decentralized training of deep learning models is a key element for enabling data privacy and on-device learning over networks. In realistic learning scenarios, the presence of heterogeneity across different clients’ local datasets poses an optimization challenge and may severely deteriorate the generalization performance. In this paper, we investigate and identify the limitation of several decentralized optimization algorithms for different degrees of data heterogeneity. We propose a novel momentum-based method to mitigate this decentralized training difficulty. We show in extensive empirical experiments on various CV/NLP datasets (CIFAR-10, ImageNet, and AG News) and several network topologies (Ring and Social Network) that our method is much more robust to the heterogeneity of clients’ data than other existing methods, by a significant improvement in test performance (1%−20%). Our code is publicly available1. C3 - Proceedings of the 38th International Conference on Machine Learning DA - 2021/// PY - 2021 VL - 139 SP - 6654 EP - 6665 LA - en PB - PMLR UR - https://proceedings.mlr.press/v139/lin21c.html AN - http://infoscience.epfl.ch/record/288750 DB - Infoscience ER -