We present an extension of the Nearest Neighbour classifier that can adapt to sample imbalances in local regions of the dataset. Our approach uses the hubness statistic as a measure of a relation between new samples and the existing training set. This allows to estimate the upper limit of neighbours that vote for the label of the new instance. This estimation improves the classifier performance in situations where some classes are locally under-represented. The main focus of our method is to solve the problem of local undersampling that exists in hyperspectral data classification. Using several well-known Machine Learning and hyperspectral datasets, we show that our approach outperforms standard and distance-weighted kNN, especially for high values of k.
a978-3-030-00840-6 01416nas a2200145 4500000000100000000000100001008004100002260004400043100002400087700002100111700002100132245009200153520100300245020002201248 2018 d bSpringer International Publishing aCham 1 aMichał Romaszewski 1 aArkadiusz Sochan 1 aKrzysztof Skabek 00 aMatrix and Tensor-Based Approximation of 3D Face Animations from Low-Cost Range Sensors 3 aThree-dimensional animation is often represented in the form of a sequence of 3D meshes, also called dynamic animation or Temporally Coherent Mesh Sequence (TCMS). Widespread availability of affordable range sensors makes capturing such data easy, however, its huge volume complicates both storage and further processing. One of the possible solutions is to approximate the data using matrix or tensor decomposition. However the quality the animation may have different impact on both approaches. In this work we use the Microsoft Kinect™ to crate sequences of human face models and compare the approximation error obtained from modelling animations using Principal component analysis (PCA) and Higher Order Singular Value Decomposition (HOSVD). We focus on distortion introduced by reconstruction of data from its truncated factorization. We show that while HOSVD may outperform PCA in terms of approximation error, it may be significantly affected by distortion in animation data.
a978-3-030-00840-6 00427nas a2200157 4500000000100000000000100001008004100002260001000043100002000053700002300073700002400096245011600120300001200236490000700248022001400255 2019 d cMarch 1 aMichał Cholewa 1 aPrzemysław Głomb 1 aMichał Romaszewski 00 aA Spatial-Spectral Disagreement-Based Sample Selection With an Application to Hyperspectral Data Classification a467-471 0 v16 a1545-598X 00555nas a2200169 4500000000100000000000100001008004100002260007300043100002300116700002100139700002400160700002000184245011100204856003700315300001100352020002200363 2019 d c2019 bWroclaw University of Science and Technology aWrocław, Poland 1 aPrzemysław Głomb 1 aKrzysztof Domino 1 aMichał Romaszewski 1 aMichał Cholewa 00 aBand selection with Higher Order Multivariate Cumulants for small target detection in hyperspectral images uhttps://arxiv.org/abs/1808.03513 ap. 121 a978-83-943803-2-8 00446nas a2200145 4500000000100000000000100001008004100002260006700043100002200110700002300132700002400155700002400179245007500203020002200278 2019 d bWroclaw University of Science and Technology aWrocław, Poland 1 aBartosz Grabowski 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aMateusz Ostaszewski 00 aUnsupervised deep learning approach to hyperspectral anomaly detection a978-83-943803-2-8 00409nas a2200133 4500000000100000000000100001008004100002260001300043100002300056700002400079245006100103856008900164020002200253 2020 d bElsevier 1 aPrzemysław Głomb 1 aMichał Romaszewski 00 aAnomaly detection in hyperspectral remote sensing images uhttps://www.elsevier.com/books/hyperspectral-remote-sensing/pandey/978-0-08-102894-0 a978-0-08-102894-0 03407nas a2200193 4500000000100000000000100001008004100002260001200043100002000055700002400075700002300099700002200122700002000144245008300164856004600247490000700293520289900300022001403199 2020 d c11/2020 1 aKamil Książek 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aBartosz Grabowski 1 aMichał Cholewa 00 aBlood Stain Classification with Hyperspectral Imaging and Deep Neural Networks uhttps://www.mdpi.com/1424-8220/20/22/6666 0 v20 3 aIn recent years, growing interest in deep learning neural networks has raised a question on how they can be used for effective processing of high-dimensional datasets produced by hyperspectral imaging (HSI). HSI, traditionally viewed as being within the scope of remote sensing, is used in non-invasive substance classification. One of the areas of potential application is forensic science, where substance classification on the scenes is important. An example problem from that area—blood stain classification—is a case study for the evaluation of methods that process hyperspectral data. To investigate the deep learning classification performance for this problem we have performed experiments on a dataset which has not been previously tested using this kind of model. This dataset consists of several images with blood and blood-like substances like ketchup, tomato concentrate, artificial blood, etc. To test both the classic approach to hyperspectral classification and a more realistic application-oriented scenario, we have prepared two different sets of experiments. In the first one, Hyperspectral Transductive Classification (HTC), both a training and a test set come from the same image. In the second one, Hyperspectral Inductive Classification (HIC), a test set is derived from a different image, which is more challenging for classifiers but more useful from the point of view of forensic investigators. We conducted the study using several architectures like 1D, 2D and 3D convolutional neural networks (CNN), a recurrent neural network (RNN) and a multilayer perceptron (MLP). The performance of the models was compared with baseline results of Support Vector Machine (SVM). We have also presented a model evaluation method based on t-SNE and confusion matrix analysis that allows us to detect and eliminate some cases of model undertraining. Our results show that in the transductive case, all models, including the MLP and the SVM, have comparative performance, with no clear advantage of deep learning models. The Overall Accuracy range across all models is 98–100% for the easier image set, and 74–94% for the more difficult one. However, in a more challenging inductive case, selected deep learning architectures offer a significant advantage; their best Overall Accuracy is in the range of 57–71%, improving the baseline set by the non-deep models by up to 9 percentage points. We have presented a detailed analysis of results and a discussion, including a summary of conclusions for each tested architecture. An analysis of per-class errors shows that the score for each class is highly model-dependent. Considering this and the fact that the best performing models come from two different architecture families (3D CNN and RNN), our results suggest that tailoring the deep neural network architecture to hyperspectral data is still an open problem.
a1424-8220 00422nas a2200157 4500000000100000000000100001008004100002260001200043100002300055700002400078700002100102700002300123245010100146300001000247490000700257 2020 d c11/2020 1 aMałgorzata Domino 1 aMichał Romaszewski 1 aTomasz Jasiński 1 aMałgorzata Maśko 00 aComparison of the Surface Thermal Patterns of Horses and Donkeys in Infrared Thermography Images a 2201 0 v10 01424nas a2200145 4500000000100000000000100001008004100002100002400043700002300067700002100090700002000111245008000131856004300211520102400254 2020 d 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aArkadiusz Sochan 1 aMichał Cholewa 00 aA Dataset for Evaluating Blood Detection in Hyperspectral Images [Data set] uhttps://doi.org/10.5281/zenodo.3984905 3 aThe sensitivity of hyperspectral imaging (imaging spectroscopy) to haemoglobin derivatives makes it a promising tool for detection and classification of blood. However, due to complexity and high dimensionality of hyperspectral images, the development of hyperspectral blood detection algorithms is challenging. To facilitate their development, we present a new hyperspectral blood detection dataset. This dataset consists of 14 hyperspectral images (ENVI format) of a mock-up scene containing blood and visually similar substances (e.g. artificial blood or tomato concentrate). Images were taken over a period of three weeks and differ in terms of background composition and lighting intensity. To facilitate the use of data, the dataset includes an annotation of classes: pixels where blood and similar substances are visible have been marked by the authors. The main intention behind the dataset is to serve as testing data for Machine Learning methods for hyperspectral target detection and classification.
01436nas a2200169 4500000000100000000000100001008004100002100002400043700002300067700002100090700002000111245006900131856003700200490000800237520100700245022001401252 2021 d 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aArkadiusz Sochan 1 aMichał Cholewa 00 aA dataset for evaluating blood detection in hyperspectral images uhttps://arxiv.org/abs/2008.10254 0 v320 3 aThe sensitivity of imaging spectroscopy to haemoglobin derivatives makes it a promising tool for detecting blood. However, due to complexity and high dimensionality of hyperspectral images, the development of hyperspectral blood detection algorithms is challenging. To facilitate their development, we present a new hyperspectral blood detection dataset. This dataset, published under an open access license, consists of multiple detection scenarios with varying levels of complexity. It allows to test the performance of Machine Learning methods in relation to different acquisition environments, types of background, age of blood and presence of other blood-like substances. We have explored the dataset with blood detection experiments, for which we have used a hyperspectral target detection algorithm based on the well-known Matched Filter detector. Our results and their discussion highlight the challenges of blood detection in hyperspectral data and form a reference for further works.
a0379-0738 00772nas a2200145 4500000000100000000000100001008004100002100002300043700002400066700002100090700002300111245009000134856003800224520036400262 2020 d 1 aMałgorzata Domino 1 aMichał Romaszewski 1 aTomasz Jasiński 1 aMałgorzata Maśko 00 aComparison of surface thermal patterns of horses and donkeys in IRT images [Data set] uhttps://zenodo.org/record/4085075 3 aThe dataset contains infrared thermography (IRT) images of horses and donkeys along with annotations of 15 ROIs corresponding to areas with potential impact of large equine muscles. All animals were healthy and rested. Imaged animals: sixteen horses and eighteen donkeys. Measured animal features such as hair length, and skin thickness are included.
00567nas a2200181 4500000000100000000000100001008004100002260001200043100001700055700002300072700001800095700002400113700002000137245016200157856004500319490000700364022001400371 2021 d c03/2021 1 aFilip Pałka 1 aWojciech Książek 1 aPawel Plawiak 1 aMichał Romaszewski 1 aKamil Książek 00 aHyperspectral Classification of Blood-Like Substances Using Machine Learning Methods Combined with Genetic Algorithms in Transductive and Inductive Scenarios uhttps://www.mdpi.com/1424-8220/21/7/2293 0 v21 a1424-8220 01481nas a2200205 4500000000100000000000100001008004100002260004200043100002000085700002300105700002400128700002000152700002200172700001900194245010400213856003700317490001000354520088900364022002201253 2022 d c05/2022 bSpringer, Cham aLecce, Italy 1 aKamil Książek 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aMichał Cholewa 1 aBartosz Grabowski 1 aKrisztian Buza 00 aImproving Autoencoder Training Performance for Hyperspectral Unmixing with Network Reinitialisation uhttps://arxiv.org/abs/2109.13748 0 v13231 3 aNeural networks, in particular autoencoders, are one of the most promising solutions for unmixing hyperspectral data, i.e. reconstructing the spectra of observed substances (endmembers) and their relative mixing fractions (abundances), which is needed for effective hyperspectral analysis and classification. However, as we show in this paper, the training of autoencoders for unmixing is highly dependent on weights initialisation; some sets of weights lead to degenerate or low-performance solutions, introducing negative bias in the expected performance. In this work, we experimentally investigate autoencoders stability as well as network reinitialisation methods based on coefficients of neurons’ dead activations. We demonstrate that the proposed techniques have a positive effect on autoencoder training in terms of reconstruction, abundances and endmembers errors.
a978-3-031-06427-2 01389nas a2200193 4500000000100000000000100001008004100002260001200043100002300055700002000078700001900098700001800117700002400135245006600159856004000225490000700265520090900272022001401181 2023 d c05/2023 1 aPrzemysław Głomb 1 aMichał Cholewa 1 aWojciech Koral 1 aAndrzej Madej 1 aMichał Romaszewski 00 aDetection of emergent leaks using machine learning approaches uhttps://doi.org/10.2166/ws.2023.118 0 v23 3 aIn this work, we focus on the detection of leaks occurring in district metered areas (DMAs). Those leaks are observable as a number of time-related deviations from zone patterns over days or weeks. While they are detectable given enough time, due to the huge cost of water loss resulting from an undetected leak, the main challenge is to find them as soon as possible, when the deviation from the zone pattern is small. Using our collected observational data, we investigate the appearance of leaks and discuss the performance of several machine learning (ML) anomaly detectors in detecting them. We test a diverse set of six anomaly detectors, each based on a different ML algorithm, on nine scenarios containing leaks and anomalies of various kinds. The proposed approach is very effective at quickly (within hours) identifying the presence of a leak, with a limited number of false positives.
a1606-9749 02406nas a2200241 4500000000100000000000100001008004100002260001200043100002000055700002200075700002300097700002400120700001800144700001900162700001900181700001700200700001900217700001900236700002100255245017800276856006600454520164400520 2023 d c08/2023 1 aKamil Książek 1 aWilhelm Masarczyk 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aIga Stokłosa 1 aPiotr Ścisło 1 aPaweł Dębski 1 aRobert Pudlo 1 aKrisztian Buza 1 aPiotr Gorczyca 1 aMagdalena Piegza 00 aThe analysis of heart rate variability and accelerometer mobility data in the assessment of symptom severity in psychotic disorder patients using a wearable Polar H10 sensor uhttps://www.medrxiv.org/content/10.1101/2023.08.04.23293640v1 3 aBackground and Objective: Advancement in mental health care requires easily accessible diagnostic and treatment assessment tools. There is an ongoing search for biomarkers that would enable objectification and automatization of the diagnostic and treatment process dependent on a psychiatric interview. Current wearable technology and computational methods make it possible to incorporate heart rate variability (HRV), an indicator of autonomic nervous system functioning and a potential biomarker of disease severity in mental disorders, into accessible diagnostic and treatment assessment frameworks.
Methods: We used a commercially available electrocardiography (ECG) chest strap with a built-in accelerometer, i.e. Polar H10, to record R-R intervals and activity of 30 hospitalized schizophrenia or bipolar disorder patients and 30 control participants for 1.5-2 hours time periods.
We performed an analysis to assess the relationship between HRV and the Positive and Negative Syndrome Scale (PANSS) test scores. The source code for the reproduction of all experiments is available on GitHub while the dataset is available in Zenodo.
Results and Conclusions: Mean HRV values were lower in the patient group and negatively correlated with the results of the PANSS general subcategory. For the control group, we also discovered the inversely proportional dependency between the mobility coefficient based on accelerometer data and HRV. This relationship was less pronounced for the patient group. This indicates that HRV and mobility may be promising markers in disease diagnosis.
The issue of diagnosing psychotic diseases, including schizophrenia and bipolar disorder, in particular, the objectification of symptom severity assessment, is still a problem requiring the attention of researchers. Two measures that can be helpful in patient diagnosis are heart rate variability calculated based on electrocardiographic signal and accelerometer mobility data. The following dataset contains data from 30 psychiatric ward patients having schizophrenia or bipolar disorder and 30 healthy persons. The duration of the measurements for individuals was usually between 1.5 and 2 hours. R-R intervals necessary for heart rate variability calculation were collected simultaneously with accelerometer data using a wearable Polar H10 device. The Positive and Negative Syndrome Scale (PANSS) test was performed for each patient participating in the experiment, and its results were attached to the dataset. Furthermore, the code for loading and preprocessing data, as well as for statistical analysis, was included on the corresponding GitHub repository.
00488nas a2200157 4500000000100000000000100001008004100002100002400043700002100067700002400088700002300112700002600135700002000161245013500181022001400316 2023 d 1 aAgnieszka Ciemięga 1 aKatarzyna Maresz 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aPaulina Krupska-Wolas 1 aKrystian Prusik 00 aHyperspectral Imaging as a Facile and Non-Destructive Method for Size Analysis of Gold Nanoparticles Deposited on Porous Materials a0934-0866 00491nas a2200169 4500000000100000000000100001008004100002260004000043100002300083700002400106700002000130700001900150700001800169700002000187700002500207245008900232 2023 d bWydawnictwo Politechniki Łódzkiej 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aMichał Cholewa 1 aWojciech Koral 1 aAndrzej Madej 1 aMaciej Skrabski 1 aKatarzyna Kołodziej 00 aMachine Learning for Water Leak Detection and Localization in the WaterPrime Project 00754nas a2200253 4500000000100000000000100001008004100002260001200043100002000055700002200075700002300097700002400120700001800144700001900162700001900181700001700200700001900217700001900236700002100255245013000276856007200406490000800478022001400486 2024 d c05/2024 1 aKamil Książek 1 aWilhelm Masarczyk 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aIga Stokłosa 1 aPiotr Ścisło 1 aPaweł Dębski 1 aRobert Pudlo 1 aKrisztian Buza 1 aPiotr Gorczyca 1 aMagdalena Piegza 00 aAssessment of symptom severity in psychotic disorder patients based on heart rate variability and accelerometer mobility data uhttps://www.sciencedirect.com/science/article/pii/S0010482524006280 0 v176 a0010-4825 00457nas a2200157 4500000000100000000000100001008004100002260001200043100002400055700002400079700002300103700002000126700002500146245009100171856003700262 2024 d c06/2024 1 aMichał Romaszewski 1 aPrzemysław Sekuła 1 aPrzemysław Głomb 1 aMichał Cholewa 1 aKatarzyna Kołodziej 00 aThrough the Thicket: A Study of Number-Oriented LLMs derived from Random Forest Models uhttps://arxiv.org/abs/2406.04926 00578nas a2200193 4500000000100000000000100001008004100002260001200043100002000055700002400075700002300099700002500122700002100147700001600168700001900184700001800203700002000221245014300241 2024 d c06/2024 1 aMichał Cholewa 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aKatarzyna Kołodziej 1 aMichał Gorawski 1 aJakub Koral 1 aWojciech Koral 1 aAndrzej Madej 1 aKryspin Musioł 00 a‘Just One More Sensor is Enough’ – Iterative Water Leak Localization with Physical Simulation and a Small Number of Pressure Sensors 00456nas a2200157 4500000000100000000000100001008004100002260001200043100002500055700002000080700002300100700001900123700002400142245009500166856003700261 2024 d c09/2024 1 aKatarzyna Kołodziej 1 aMichał Cholewa 1 aPrzemysław Głomb 1 aWojciech Koral 1 aMichał Romaszewski 00 aEfficient Numerical Calibration of Water Delivery Network Using Short-Burst Hydrant Trials uhttps://arxiv.org/abs/2410.02772 00409nas a2200145 4500000000100000000000100001008004100002100002500043700002000068700002400088700002300112700001900135700002400154245008500178 2024 d 1 aKatarzyna Kołodziej 1 aMichał Cholewa 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aWojciech Koral 1 aPrzemysław Sekuła 00 aCalibration of Water Demand Network Through a Machine Learned Digital Twin Model 00305nas a2200109 4500000000100000000000100001008004100002260002800043100002400071700002400095245007600119 2024 d c04/2024 aWarsaw, Poland 1 aMichał Romaszewski 1 aPrzemysław Sekuła 00 aPoster: Explainable classification of multimodal time series using LLMs 00427nas a2200157 4500000000100000000000100001008004100002260001200043100002400055700002400079700002300103700002000126700002500146245009100171490000700262 2025 d c03/2025 1 aMichał Romaszewski 1 aPrzemysław Sekuła 1 aPrzemysław Głomb 1 aMichał Cholewa 1 aKatarzyna Kołodziej 00 aThrough the Thicket: A Study of Number-Oriented LLMS Derived from Random Forest Models 0 v15 01940nas a2200157 4500000000100000000000100001008004100002100002400043700002400067700002300091700002000114700001900134245006100153856006700214520150100281 2025 d 1 aPrzemysław Sekuła 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aMichał Cholewa 1 aŁukasz Pawela 00 aQuantum-aware Transformer model for state classification uhttps://link.springer.com/chapter/10.1007/978-3-031-97570-7_15 3 aEntanglement is a fundamental feature of quantum mechanics, playing a crucial role in quantum information processing. However, classifying entangled states, particularly in the mixed-state regime, remains a challenging problem, especially as system dimensions increase. In this work, we focus on bipartite quantum states and present a data-driven approach to entanglement classification using transformer-based neural networks. Our dataset consists of a diverse set of bipartite states, including pure separable states, Werner entangled states, general entangled states, and maximally entangled states. We pretrain the transformer in an unsupervised fashion by masking elements of vectorized Hermitian matrix representations of quantum states, allowing the model to learn structural properties of quantum density matrices. This approach enables the model to generalize entanglement characteristics across different classes of states. Once trained, our method achieves near-perfect classification accuracy, effectively distinguishing between separable and entangled states. Compared to previous Machine Learning, our method successfully adapts transformers for quantum state analysis, demonstrating their ability to systematically identify entanglement in bipartite systems. These results highlight the potential of modern machine learning techniques in automating entanglement detection and classification, bridging the gap between quantum information theory and artificial intelligence.
00357nas a2200121 4500000000100000000000100001008004100002100002000043700002400063700002000087245010700107856002100214 d 1 aKatarzyna Filus 1 aMichał Romaszewski 1 aMateusz Żarski 00 aSemantic Depth Matters: Explaining Errors of Deep Vision Networks through Perceived Class Similarities uarXiv:2504.09956 00463nas a2200157 4500000000100000000000100001008004100002260001200043100001700055700002400072700001700096245009400113856007700207490000700284022001400291 2025 d c08/2025 1 aManish Gupta 1 aMichał Romaszewski 1 aPiotr Gawron 00 aPotential of quantum machine learning for processing multispectral Earth observation data uhttps://journals.pan.pl/dlibra/publication/154279/edition/135180/content 0 v73 a2300-1917 02630nas a2200241 4500000000100000000000100001008004100002260001200043100002000055700002200075700002300097700002400120700001900144700002400163700002000187700002500207700001900232700002100251245012200272490000700394520197300401022001402374 2025 d c09/2025 1 aKamil Książek 1 aWilhelm Masarczyk 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aKrisztian Buza 1 aPrzemysław Sekuła 1 aMichał Cholewa 1 aKatarzyna Kołodziej 1 aPiotr Gorczyca 1 aMagdalena Piegza 00 aDeep learning approach for automatic assessment of schizophrenia and bipolar disorder in patients using R-R intervals 0 v21 3 aSchizophrenia and bipolar disorder are severe mental illnesses that significantly impact quality of life. These disorders are associated with autonomic nervous system dysfunction, which can be assessed through heart activity analysis. Heart rate variability (HRV) has shown promise as a potential biomarker for diagnostic support and early screening of those conditions. This study aims to develop and evaluate an automated classification method for schizophrenia and bipolar disorder using short-duration electrocardiogram (ECG) signals recorded with a low-cost wearable device. We conducted classification experiments using machine learning techniques to analyze R-R interval windows extracted from short ECG recordings. The study included 60 participants—30 individuals diagnosed with schizophrenia or bipolar disorder and 30 control subjects. We evaluated multiple machine learning models, including Support Vector Machines, XGBoost, multilayer perceptrons, Gated Recurrent Units, and ensemble methods. Two time window lengths (about 1 and 5 minutes) were evaluated. Performance was assessed using 5-fold cross-validation and leave-one-out cross-validation, with hyperparameter optimization and patient-level classification based on individual window decisions. Our method achieved classification accuracy of 83% for the 5-fold cross-validation and 80% for the leave-one-out scenario. Despite the complexity of our scenario, which mirrors real-world clinical settings, the proposed approach yielded performance comparable to advanced diagnostic methods reported in the literature. The results highlight the potential of short-duration HRV analysis as a cost-effective and accessible tool for aiding in the diagnosis of schizophrenia and bipolar disorder. Our findings support the feasibility of using wearable ECG devices and machine learning-based classification for psychiatric screening, paving the way for further research and clinical applications.
a1553-7358 00317nas a2200109 4500000000100000000000100001008004100002100002000043700002400063700002300087245009700110 2025 d 1 aMichał Cholewa 1 aMichał Romaszewski 1 aPrzemysław Głomb 00 aData structure better than labels? Unsupervised heuristics for SVM hyperparameter estimation 00317nas a2200109 4500000000100000000000100001008004100002100002000043700002400063700002300087245009700110 2025 d 1 aMichał Cholewa 1 aMichał Romaszewski 1 aPrzemysław Głomb 00 aData structure better than labels? Unsupervised heuristics for SVM hyperparameter estimation 00485nas a2200169 4500000000100000000000100001008004100002260001200043100002500055700002000080700002300100700001900123700002400142245009500166856004600261490000800307 2026 d c02/2026 1 aKatarzyna Kołodziej 1 aMichał Cholewa 1 aPrzemysław Głomb 1 aWojciech Koral 1 aMichał Romaszewski 00 aEfficient Numerical Calibration of Water Delivery Network Using Short-Burst Hydrant Trials uhttps://doi.org/10.1061/JWRMD5.WRENG-6839 0 v152 00394nas a2200145 4500000000100000000000100001008004100002260001200043100002200055700002400077700002000101700002100121700001900142245008700161 2025 d c04/2025 1 aKrzysztof Grochla 1 aMichał Romaszewski 1 aJarosław Sotor 1 aPiotr Masłowski 1 aBartosz Belter 00 aTowards AI-assisted Science with PLAI4SCIENCE - new Polish Research Infrastructure 00365nas a2200145 4500000000100000000000100001008004100002100001800043700002000061700001700081700002300098700002400121700002000145245005400165 2026 d 1 aAnna Zawadzka 1 aMateusz Żarski 1 aKamil Drejer 1 aPrzemysław Głomb 1 aMichał Romaszewski 1 aMichał Cholewa 00 a Map-Guided Cross-Training for Building Detection 02216nas a2200169 4500000000100000008004100001100001700042700002500059700002000084700002400104700002300128700001800151700001600169700001900185245007600204520176600280 2026 d 1 aAygün Varol 1 aKatarzyna Kołodziej 1 aŁukasz Sobczak 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aNaser Motlagh 1 aMirka Leino 1 aJohanna Virkki 00 aEnabling Cloud-Level Accuracy in Edge AI through IoT Data Preprocessing 3 aLarge language models (LLMs) offer a natural-language interface for interpreting Internet of Things (IoT) sensor data in smart environments; however, cloud deployment introduces latency, privacy, and connectivity concerns. Local LLMs can reduce these limitations, but compact edge-deployable models often show weaker numerical reasoning when raw sensor readings are provided directly. This paper investigates whether prompt-side preprocessing can improve the accuracy--latency trade-off of local LLMs for environmental monitoring. We propose a structured prompt construction framework that transforms raw air-quality and thermal-comfort measurements into progressively enriched textual representations: raw sensor values, threshold-aware descriptions, and compact environmental summary flags. The approach is evaluated using indoor Raspberry Pi/BME680 datasets from Tampere University and outdoor air-quality datasets from Helsinki, Katowice, and Warsaw. We construct a binary LLM query dataset covering air quality, thermal comfort, and joint environmental conditions, and evaluate five local and five cloud LLMs across three prompt variants and two inference modes, with and without chain-of-thought prompting. The results show that prompt enrichment substantially improves local-model accuracy. In No-CoT mode, local accuracy increases from 50.9\% to 81.7\% indoors and from 63.7\% to 89.3\% outdoors from the raw to the most enriched prompt. Local No-CoT inference is the fastest configuration, with mean latency close to 0.22 s, while CoT substantially increases inference time. These findings suggest that lightweight prompt-side preprocessing can narrow the local--cloud performance gap and support low-latency IoT analytics in smart environments.
02266nas a2200181 4500000000100000008004100001100001700042700002500059700002000084700002400104700002300128700001800151700001600169700001900185245007600204856003800280520176600318 2026 d 1 aAygün Varol 1 aKatarzyna Kołodziej 1 aŁukasz Sobczak 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aNaser Motlagh 1 aMirka Leino 1 aJohanna Virkki 00 aEnabling Cloud-Level Accuracy in Edge AI through IoT Data Preprocessing uhttps://arxiv.org/abs/2606.22496 3 aLarge language models (LLMs) offer a natural-language interface for interpreting Internet of Things (IoT) sensor data in smart environments; however, cloud deployment introduces latency, privacy, and connectivity concerns. Local LLMs can reduce these limitations, but compact edge-deployable models often show weaker numerical reasoning when raw sensor readings are provided directly. This paper investigates whether prompt-side preprocessing can improve the accuracy--latency trade-off of local LLMs for environmental monitoring. We propose a structured prompt construction framework that transforms raw air-quality and thermal-comfort measurements into progressively enriched textual representations: raw sensor values, threshold-aware descriptions, and compact environmental summary flags. The approach is evaluated using indoor Raspberry Pi/BME680 datasets from Tampere University and outdoor air-quality datasets from Helsinki, Katowice, and Warsaw. We construct a binary LLM query dataset covering air quality, thermal comfort, and joint environmental conditions, and evaluate five local and five cloud LLMs across three prompt variants and two inference modes, with and without chain-of-thought prompting. The results show that prompt enrichment substantially improves local-model accuracy. In No-CoT mode, local accuracy increases from 50.9\% to 81.7\% indoors and from 63.7\% to 89.3\% outdoors from the raw to the most enriched prompt. Local No-CoT inference is the fastest configuration, with mean latency close to 0.22 s, while CoT substantially increases inference time. These findings suggest that lightweight prompt-side preprocessing can narrow the local--cloud performance gap and support low-latency IoT analytics in smart environments.
03010nas a2200181 4500000000100000008004100001100001700042700002500059700002000084700002400104700002300128700001800151700001600169700001900185245011600204856005200320520245600372 2026 d 1 aAygün Varol 1 aKatarzyna Kołodziej 1 aŁukasz Sobczak 1 aMichał Romaszewski 1 aPrzemysław Głomb 1 aNaser Motlagh 1 aMirka Leino 1 aJohanna Virkki 00 aAIRWISE: Environmental Sensor and LLM Query Dataset for Edge AI-Based Air-Quality and Thermal-Comfort Analytics uhttps://zenodo.org/doi/10.5281/zenodo.20783874 3 aAIRWISE is an open benchmark dataset for edge AI, IoT environmental monitoring, and large language model (LLM) evaluation on real sensor data. It combines indoor air-quality and thermal-comfort measurements, outdoor air-quality and meteorological time series, and a labeled binary question-answering benchmark for threshold-aware reasoning and anomaly detection. The dataset supports the study “Enabling Cloud-Level Accuracy in Edge AI through IoT Data Preprocessing”: https://doi.org/10.48550/arXiv.2606.22496
AIRWISE provides three core resources:
Indoor environmental sensor dataset (BME680, 1-minute resolution). 146,000+ minute-averaged readings from Raspberry Pi + Bosch BME680 nodes in three real micro-environments (office, kitchen, hallway) at Tampere, Finland (2025-11-14 to 2025-12-18): temperature, relative humidity, barometric pressure, gas resistance, indoor air quality (IAQ) proxy, z-score anomaly detectors, and limit-annotated long-format tables with ground-truth anomaly flags.
Outdoor air-quality and meteorological time series (hourly, full-year 2023). Helsinki (Finland), Katowice and Warsaw (Poland): NO₂, CO, O₃, PM₂.₅, temperature, and relative humidity assembled from national monitoring networks (FMI, GIOŚ, IMGW-PIB), as cleaned single-series city tables (≈8,760 hourly records per city) and limit-annotated long-format tables with anomaly flags.
Binary LLM query benchmark (1,440 labeled yes/no questions). 240 natural-language questions per site across six sites, balanced by task type (80 air quality, 80 thermal comfort, 80 joint environmental condition), each with ground-truth answer, violation label, offending factors, and rationale — ready for benchmarking local/edge and cloud LLMs on sensor-grounded reasoning.
Also included: reference air-quality and thermal-comfort limits (EU air-quality standards and Finnish/Polish national guidelines); the unmodified institutional source files for the outdoor component (FMI exports, GIOŚ archive workbooks and station metadata, IMGW-PIB synoptic tables); the complete dataset-generation scripts covering every pipeline step (raw sources → cleaned tables → limit-annotated data → LLM queries) with a requirements.txt for a verified minimal Python environment; SHA-256 checksums; and a dependency-free validation script. Schemas, exact row counts, coverage, and usage examples are documented in README.md and MANIFEST.md.