9,909 research outputs found
Metadata Representations for Queryable ML Model Zoos
No full textMachine learning (ML) practitioners and organizations are building model zoos of pre-trained models, containing metadata describing properties of the ML models and datasets that are useful for reporting, auditing, reproducibility, and interpretability purposes. The metatada is currently not standardised; its expressivity is limited; and there is no interoperable way to store and query it. Consequently, model search, reuse, comparison, and composition are hindered. In this paper, we advocate for standardized ML model metadata representation and management, proposing a toolkit supported to help practitioners manage and query that metadata.Web Information SystemsHuman-Centred Artificial Intelligenc
A Manifesto of Nodalism
No full textThis paper proposes the notion of Nodalism as a means describing contemporary culture and of understanding my own creative practice in electronic music composition. It draws on theories and ideas from Kirby, Bauman, Bourriaud, Deleuze, Guatarri, and Gochenour, to demonstrate how networks of ideas or connectionist neural models of cognitive behaviour can be used to contextualize, understand and become a creative tool for the creation of contemporary electronic music
K:D-Rib un inibitore della proliferazione delle cellule tumorali ed un promotore del folding di DNAzima.
No full textLe cellule tumorali mostrano un aumento della glicolisi anche in presenza d’ossigeno disponibile. Questo processo chiamato “Effetto Warburg” è uno dei fondamentali switch metabolici che una cellula cancerosa manifesta se paragonata ad una cellula non cancerosa. Tra i trasportatori del glucosio c’è la famiglia SGLT (cotrasportatore Na+/glucosio). L’elevata concentrazione totale di sodio in lesioni mammarie tumorali misurata con 23Na-MRI sembra essere un indicatore del livello cellulare di malignità [1].
Il ruolo del D-ribosio e dello ione potassio (K+) sono ampiamente noti: il D-ribosio è un aldopentoso, assiste il metabolismo energetico della cellula oltre ad esser un precursore d’alcuni aminoacidi; lo ione K+ è coinvolto in molti processi tra cui apoptosi, genesi e mantenimento del potenziale di membrana e stabilizza il folding dei G-quadruplex. C'è un interesse crescente nell’utilizzo di G-quadruplex anche come sensori per lo ione K+ [2].
Per saggiare l’effetto del K:D-Rib sulla proliferazione delle linee cellulari non si sono potuti utilizzare i comuni saggi metabolici (MTT assay, WST-1 assay ecc) in quanto K:D-Rib interagisce con il bromuro di tetrazolio riducendolo a formazan [3].
Ci siamo perciò avvalsi di un metodo d’uso comune: le cellule (4000 cell/ml trattate con K:D-Rib 5mM e incubate per 12 gg) sono state contate utilizzando il programma ImageJ ed in seguito si è calcolato il tempo di duplicazione. Il controllo ha mostrato un tempo di duplicazione di 44 ore, mentre il trattato di 59 ore. L’effetto rilevato è un progressivo rallentamento della proliferazione cellulare del trattato rispetto al controllo.
Il saggio delle colonie, nel quale si contano gruppi cellulari di almeno 50 unità, ha confermato i risultati delle curve di crescita, mostrando ancora una volta un effetto antiproliferativo del K:D-Rib sia sulla linea HTB-126 [3] che HTB-30 (metastasi pleurica di carcinoma mammario). Quest’ultima linea
66 |
cellulare ha mostrato 15 colonie per il controllo e nessuna colonia per i trattati. Questi dati sperimentali unitamente ad una ricca bibliografia sui canali K+ [2] e sul ruolo stesso del K+ ci hanno portato a ipotizzare che una fine regolazione della concentrazione di K+ stia alla base del “corretto funzionamento” delle cellule. Da qui l’esigenza di valutare se K+ entri o no nelle cellule in seguito al trattamento con K:D-Rib. La concentrazione di K+ è stata misurata sfruttando la formazione di una macromolecola chiamata DNAzima [4]. Dalla concentrazione di DNAzima si è in grado, attraverso spettroscopia UV-VIS, di misurare se lo ione K+ è presente nella soluzione. Abbiamo quindi verificato la formazione di questa molecola in presenza di K:D-Rib 5mM. Il DNAzima così formato può essere utilizzato come biosensore per la misura della concentrazione del K+ nel mezzo cellulare, prima e dopo il trattamento. Risultati preliminari sembrano indicare che il surnatante delle HTB-126 trattate con K:D-Rib 5mM per 48h mostrano una concentrazione di DNAzima inferiore rispetto al DMEM (terreno di coltura delle cellule) con K:D-Rib 5mM, ma superiore al solo DMEM ed al surnatante del controllo (HTB-126 non trattate). Possiamo dire che una parte di K+ entri all’interno della cellula e che il DNAzima possa essere utilizzato come biosensore di K+.
Bibliografia:
[1] Ronald Ouwerkerk, Michael A. Jacobs , Katarzyna J. Macura, Antonio C. Wolff, Vered Stearn, Sarah D. Mezban, Nagi F. Khouri, David A. Bluemke, Paul A. Bottomley Elevated tissue sodium concentration in malignant breast lesions detected with non-invasive 23Na MRI. Breast Cancer Res Treat (2007) 106:151–160.
[2] Zhiguo Wang Roles of K+ channels in regulating tumour cell proliferation and apoptosis. Eur J Physiol (2004) 448: 274–28:
[3] S. Croci, L. Bruni, S. Bussolati, M. Castaldo, M. Dondi Potassium bicarbonate and D-ribose effects on A72 canine and HTB-126 human cancer cell line proliferation in vitro. Cancer Cell International (2011) vol. 11 11-30.
[4] Travascio P, Li Y, Sen D.. DNA-enhanced peroxidase activity of a DNA-aptamer-hemin complex. Chem Biol (1998) 5:5 05–517
ZFP423 Coordinates Notch and Bone Morphogenetic Protein Signaling, Selectively Up-regulating Hes5 Gene Expression
No full textZinc finger protein 423 encodes a 30 Zn-finger transcription factor involved in cerebellar and olfactory development. ZFP423 is a known interactor of SMAD1-SMAD4 and of Collier/Olf-1/EBF proteins, and acts as a modifier of retinoic acid-induced differentiation. In the present article, we show that ZFP423 interacts with the Notch1 intracellular domain in mammalian cell lines and in Xenopus neurula embryos, to activate the expression of the Notch1 target Hes5/ESR1. This effect is antagonized by EBF transcription factors, both in cultured cells and in Xenopus embryos, and amplified in vitro by BMP4, suggesting that ZFP423 acts to integrate BMP and Notch signaling, selectively promoting their convergence onto the Hes5 gene promoter
Optimizing ML Inference Queries Under Constraints
No full textThe proliferation of pre-trained ML models in public Web-based model zoos facilitates the engineering of ML pipelines to address complex inference queries over datasets and streams of unstructured content. Constructing optimal plan for a query is hard, especially when constraints (e.g. accuracy or execution time) must be taken into consideration, and the complexity of the inference query increases. To address this issue, we propose a method for optimizing ML inference queries that selects the most suitable ML models to use, as well as the order in which those models are executed. We formally define the constraint-based ML inference query optimization problem, formulate it as a Mixed Integer Programming (MIP) problem, and develop an optimizer that maximizes accuracy given constraints. This optimizer is capable of navigating a large search space to identify optimal query plans on various model zoos.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Web Information SystemsHuman-Centred Artificial Intelligenc
Upscaling and downscaling approaches for early season rice yield prediction using Sentinel-2 and machine learning for precision nitrogen fertilisation
No full textEarly season yield prediction could support rice farmers in adopting precision agriculture for nitrogen fertilisation management. Remote sensing and machine learning (ML) can be used to predict and map crop yield during phenological stages relevant to nitrogen application, like tillering in rice, at both within-field and field scales. This study evaluated the transferability of ML models in early season yield prediction through upscaling and downscaling approaches. The effects of two prediction times (tillering and ripening stages) and training/ testing set sizes on ML models performance were evaluated over five rice growing seasons (from 2018 to 2022) in northern Italy, using whole-field-average yields and yield maps. Vegetation indices from Sentinel-2 imagery using the Google Earth Engine platform fed five ML algorithms (Cubist-CUB, Gaussian Process Regression-GPR, Neural Network-NNET, Random Forest-RF, and Support Vector Machines-SVM). ML algorithms were trained with yield maps and tested with whole-field-average yields to obtain a downscaling approach, while the opposite was done to obtain an upscaling approach. The downscaling approach showed higher accuracy than upscaling approach. Ripening stage predictions were more accurate than tillering stages, although the downscaling approach showed small differences between tillering and ripening stages. The highest tillering stage accuracy was achieved by SVM for both downscaling and upscaling approaches with 20 % and 27.8 % of Normalized Root Mean Square Error (NRMSE), and 0.99 and 0.99 of Simple Additive Weighting (SAW) score, respectively. Set size and data distribution effected ML models accuracy, with the highest performance achieved by RF and GPR with 0.80 and 1.00 of SAW score for the downscaling and upscaling approaches, respectively. This study demonstrated how ML models and downscaling approach could support rice farmers to calculate the nitrogen dose using the predicted yield at the tillering stage, enabling them to apply a site-specific nitrogen fertilisation based on the within-field yield prediction variability
REMOTE SENSING FOR THE AGRI-FOOD SECTOR: METHODS & APPLICATIONS
No full textNell'ultimo decennio, il settore agroalimentare e le autorità locali hanno investito in nuove tecnologie per affrontare le crescenti sfide sociali, economiche e ambientali legate al cibo e alla sua sostenibilità. L'intelligenza artificiale (AI) e il Machine Learning (ML) alimentate con dati ad alta risoluzione spaziale e temporale (es. telerilevamento, sensori prossimali, dati meteorologici e mappe del suolo) permettono lo sviluppo di nuovi strumenti in grado di monitorare l'intera catena agroalimentare consentendo a sua volta l'ottimizzazione dei processi produttivi e la loro sostenibilità.
In questo studio sono state indagate le tre principali applicazioni del telerilevamento per il monitoraggio delle colture: i) la mappatura delle colture, ii) la stima dei parametri biofisici e iii) la previsione delle rese. Per ognuna di queste tre applicazioni, è stato analizzato un caso studio al fine di approfondire specifici aspetti metodologici necessari allo sviluppo di un sistema di monitoraggio della filiera agroalimentare.In the last decade, the agri-food sector and local authorities have invested in new technologies to address the growing social, economic, and environmental challenges related to food and its sustainability. Artificial Intelligence (AI) and Machine Learning (ML) fed with high spatial and temporal resolution data (e.g. remote sensing, proximal sensors, weather data and soil maps) enable the development of new tools that can monitor the whole agri-food chain enabling in turn the optimization of production processes and their sustainability.
In this study, the three main applications of remote sensing for crop and land monitoring were investigated: i) crop mapping, ii) estimation of biophysical parameters, and iii) yield prediction. In particular, for each of these three applications, some methodological aspects have been analyzed to develop operational solutions for monitoring the agri-food supply chain
Assessing the temporal transferability of machine learning models for predicting processing pea yield and quality using Sentinel-2 and ERA5-land data
No full textAccurate pre-harvest prediction of yield and quality (tenderometric reading, TR) is crucial for the processing pea industry due to a narrow optimal harvest window. Machine learning (ML) models offer potential, but their real-world utility depends on their performance stability across different years. This ability, known as temporal transferability, is often overestimated by standard validation methods, creating a gap between theoretical performance and operational reliability. This study aims to rigorously quantify this temporal transferability gap for both pea yield and TR prediction. Four ML algorithms (RF, XGBoost, GPR, SVMr) were evaluated using Sentinel-2 and ERA5-Land data from 2018 to 2024 in northern Italy. A comparison was made between a standard group-based cross-validation (LOGOCV) and a temporally rigorous Leave-One-Year-Out Cross-Validation (LOYOCV). For yield prediction, ML models outperformed a baseline (NullModel) under LOGOCV (SVMr nRMSE = 18.4 %), but performance degraded significantly under LOYOCV, revealing a clear transferability gap. TR prediction was more challenging while RF showed promising results in LOGOCV (nRMSE = 22.1 %), all ML models were outperformed by the NullModel in the more realistic LOYOCV scenario. The findings highlight a critical temporal transferability gap, especially for the TR parameter, limiting the current operational readiness of standard ML models. It is recommended that future work focus on more robust approaches, such as models designed for temporal data (e.g., RNNs, Transformers) and higher-resolution data, to bridge the gap towards reliable real-world application
Influence of aging on serum phenytoin concentrations: A pharmacokinetic analysis based on therapeutic drug monitoring data.
No full textThe influence of aging on the pharmacokinetics of phenytoin at steady-state was evaluated retrospectically by comparing apparent oral clearance values (CL/F) in 75 patients aged 65-90 years (mean, 71.7 +/- 5.3 years) receiving phenytoin alone (n = 58) or in combination with phenobarbital (n = 17) and in an equal number of control patients aged 20-50 years (mean, 36.7 +/- 8.5 years) matched for gender, body weight, and comedication. All data were derived from the database of the therapeutic drug monitoring service (TDMS) of an academic neurological hospital. On average, elderly patients were found to exhibit slightly higher CL/F values compared with controls (14.6 +/- 4.7 ml h(-1) kg(-1) versus 13.1 +/- 4.2 ml h(-1) kg(-1), P < 0.05), the difference being probably related to the dose-dependent nature of phenytoin metabolism and the fact that elderly patients received lower dosages (4.4 +/- 1.1 mg kg(-1)day(-1) versus 5.3 +/- 1.1 mg kg(-1) day(-1), P < 0.001) and had lower serum phenytoin concentrations (14.1 +/- 5.7 microg ml(-1) versus 18.6 +/- 6.8 microg ml(-1), P < 0.0001). Gender and phenobarbital comedication were not found to exert any statistically significant influence on phenytoin CL/F. By contrast, in the elderly group, CL/F values were negatively correlated with age. On average, CL/F values decreased by about one-third between 65 and 85 years of age, but interindividual variability was considerable and age explained only 7.8% of the variation in CL/F in the elderly group. Overall, these findings indicate that aging is associated with a progressive decline in phenytoin clearance, presumably as a result of decreased drug metabolizing capacity. Because assessment was based on total serum phenytoin concentrations and the unbound fraction of phenytoin is known to decrease in old age, the influence of aging as quantified in this study may underestimate the magnitude of changes in the clearance of unbound, pharmacologically active drug. Based on these data, it is prudent to utilize initially smaller phenytoin dosages in old patients, and to make subsequent dose adjustments based on clinical response and serum drug level measurements. Interpretation of the latter, however, should take into account the possibility of an increase in the fraction of unbound drug
Building a generalisable ML pipeline at ING
No full textAdvances in data science have caused an increase in the use of Artificial Intelligence (AI), specifically Machine Learning (ML), throughout various fields. Not only in research but in the industry as well, has ML been receiving increasing amounts of interest. Many companies rely on ML models to increase the efficiency of existing processes or offer new services and products. The industry, however, is facing several additional challenges compared to the academic context. One of those challenges is applying the Development Operations (DevOps) model to an ML application, also referred to as MLOps. This thesis sets out to find the specific challenges that practitioners encounter while operationalising ML models. To do so, we perform a single-case case study on an ML pipeline built by the Trade & Communication Surveillance team at the ING bank. This case study consists of conducting a set of interviews and performing a manual code inspection of the pipeline. The team faces challenges ranging from having insufficient time for operationalising each ML project individually to operating in the highlyregulated fintech context. Their pipeline is able to deploy a single ML model but it does not generalise well to other projects. We present the first version of an application that mitigates these challenges. The application is able to deploy ML models to the development environment at ING and can be operated by data scientists to reduce the effort of operationalising an ML model. Computer Science | Software Technolog
- …
