Our Work

Python codes used for plotting Figures 1-5 and Supplementary Figures 1-13.    Manuscript title: Greenhouse gases overtake aerosols in driving moist heat variability in India during the summe…

Python codes used for plotting Figures 1-5 and Supplementary Figures 1-13.    Manuscript title: Greenhouse gases overtake aerosols in driving moist heat variability in India during the summer monsoon Journal: Communications Earth & Environment

Self-Organizing Maps (SOMs) are widely used in unsupervised anomaly detection due to their ability topreserve data topology and provide interpretable low-dimensional representations. However, most SOM…

Self-Organizing Maps (SOMs) are widely used in unsupervised anomaly detection due to their ability topreserve data topology and provide interpretable low-dimensional representations. However, most SOM-basedanomaly detection approaches rely almost exclusively on quantization error, which limits robustnessand fails to exploit the rich structural information encoded in trained maps. In this work, we propose anew unsupervised anomaly scoring method that combines sample-level quantization error with structuralproperties of the SOM. We introduce an algorithm that identifies dense neuronal kernels and broadercommunities using activation patterns and U-matrix values. Each neuron is then assigned a structuralanomaly score based on its distance to representative kernel and community neurons, its neighbourhooddivergence, and its activation frequency. The final anomaly score for each sample integrates this neuron-levelassessment with its quantization error. Experiments on four credit card fraud datasets show thatincorporating SOM structural characteristics substantially improves anomaly discrimination compared toquantization-only baselines. Notably, the proposed method also outperforms autoencoders—despite theirgreater model capacity—while preserving the interpretability that makes SOMs valuable in high-stakessettings such as financial fraud detection.

This repository includes results and data obtained during the internship of Amine Mahdjoub in 2024, working on the project "HYbridization and Adaptation in a Mediterranean coastal city" (HYAM, https:/…

This repository includes results and data obtained during the internship of Amine Mahdjoub in 2024, working on the project "HYbridization and Adaptation in a Mediterranean coastal city" (HYAM, https://hyam.osupytheas.fr/).  This project explores the genetic limits and exchanges between species, for germanders (plants) and gorgonians (animals), south of Marseille, in the National Park of Calanques, in a place where a well-known original and rare biodiversity is subject to major environmental stressors such as pollution remains from the industrial past. Over a period of six months, Amine collected, compiled, and formatted a large amount of data to characterize the environmental context of the study area—both terrestrial and marine—and to propose an initial attempt at modeling the distribution of the model organisms. This work received support from the french government under the France 2030 investment plan, as part of the Initiative d’Excellence d’Aix-Marseille Université – A*MIDEX (AMX-22-RE-AB-126).  

These NetCDF datasets describe potato thermal exposure across Europe using daily mean air temperature. The workflow classifies each daily temperature value into one of seven Thermal Threshold Classes …

These NetCDF datasets describe potato thermal exposure across Europe using daily mean air temperature. The workflow classifies each daily temperature value into one of seven Thermal Threshold Classes (TTCs), ranging from drastic cold to drastic hot, using potato-relevant temperature thresholds. The resulting counts represent the number of days in each class for each grid cell, year, and, where relevant, phenological stage. The regional growing-season windows used in the analysis distinguish northern, central, and southern Europe. The stage-wise datasets further split the growing season into four phenological stages, allowing thermal exposure to be interpreted during agronomically important periods rather than only as a whole-season summary.

Mazkur maqolada diversifikatsiyalash sharoitida ishlab chiqarish korxonalarini boshqarish tizimini takomillashtirishning nazariy-uslubiy va amaliy jihatlari tadqiq etilgan. Maqolada diversifikatsiyani…

Mazkur maqolada diversifikatsiyalash sharoitida ishlab chiqarish korxonalarini boshqarish tizimini takomillashtirishning nazariy-uslubiy va amaliy jihatlari tadqiq etilgan. Maqolada diversifikatsiyaning iqtisodiy mazmuni va shakllari, diversifikatsiya jarayonlarining korxona boshqaruv tizimiga ta’siri hamda boshqaruv samaradorligini baholash ko‘rsatkichlari tizimlashtirilgan. O‘zbekiston sanoatining 2024–2025-yillardagi rasmiy statistik ma’lumotlari asosida sanoat ishlab chiqarishining rivojlanish tendensiyalari tahlil qilingan. Tadqiqot natijasida diversifikatsiyalash sharoitida ishlab chiqarish korxonalarini boshqarishning strategik, tashkiliy-iqtisodiy va raqamli elementlarini integratsiyalashga asoslangan takomillashtirilgan boshqaruv modeli taklif etilgan. Shuningdek, korxonalarning diversifikatsiya darajasini baholash uchun konsentratsiya indeksi, kompleks ko‘rsatkichlar tizimi va ekonometrik modellashtirishdan foydalanish bo‘yicha ilmiy-amaliy tavsiyalar ishlab chiqilgan.

Development of the mammalian embryonic vascular system — interactive 3D model This is an interactive teaching model of how the mammalian cardiovascular system develops, from the first blood isla…

Development of the mammalian embryonic vascular system — interactive 3D model This is an interactive teaching model of how the mammalian cardiovascular system develops, from the first blood islands to the circulatory changes at birth. It is written for veterinary and biomedical students and their teachers. The dog is the reference species; switching to pig, ruminant (bovine) or horse shows where the outcome differs. The model is a single HTML file that runs in any modern web browser, with no installation. The model has four linked views. Embryo (3D): five stages, from vasculogenesis through the heart tube and looping, remodelling of the arteries and veins, and the fetal circulation, to the transition at birth and the adult remnants. Stages are timed in dog days of gestation and crown–rump length. The embryo body is translucent, so the vessels, heart, liver, lungs, yolk sac, placenta and umbilical cord are visible. Vessels form, fuse, regress and become ligaments in their developmental order. Blood-flow particles are coloured by oxygen saturation, and a species switch shows differences in the branching of the aortic arch and in the azygos system. Heart 3D: the heart develops from the straight tube through dextral looping and ballooning of the chambers to the four-chambered heart, and beats throughout. Two cut-away views open it: a frontal cut, in the same plane as the heart section, shows the cavities, septa, wall thickness and the flap valve of the foramen ovale; a cut through the arterial roots shows the aortic and pulmonary valves. Heart section (2D): an animated frontal section whose chamber outlines are slices of the 3D heart. It shows the endocardial cushions, atrial septation (septum primum and secundum, and the foramen ovale as a flap valve), ventricular septation, the atrioventricular and semilunar valves, the division of the outflow tract and the great arteries. Blood flow follows atrial and ventricular systole, and the fetal shunts close at birth. Aortic arches (2D): a ventral schematic after Rathke's diagram. It shows the craniocaudal formation and regression of the pharyngeal arch arteries, colour-coded by their definitive derivatives, together with the recurrent laryngeal nerves, the heart and blood flow. It ends with the species-specific branching of the aortic arch. Each stage carries an explanatory text with its molecular basis, species notes, clinical correlates and references. Labels sit outside the drawing with leader lines and can be hidden. The model can be rotated, panned and zoomed with a mouse, trackpad, touch or keyboard, and has settings for the heartbeat and rendering quality. Scientific basis. Morphology follows standard veterinary and medical embryology texts: Hyttel, Sinowatz & Vejlsted 2010; McGeady et al. 2017; Noden & de Lahunta 1985; Sadler 2019. Adult anatomy follows Dyce, Sack & Wensing 2010, Evans & de Lahunta 2013 and König & Liebich 2020. Chamber formation follows Christoffels et al. 2000 and Moorman & Christoffels 2003, and fetal circulation Rudolph 2009. The full reference list is inside the model. The model is schematic and not to scale; dog timings and vessel calibres are approximate. Authorship. Scientific content, design decisions and validation by the author; code generated with AI assistance (Claude, Anthropic) under the author's direction. Licensing. Scientific and pedagogical content: CC BY-NC-SA 4.0. Source code: MIT License. The three.js library (MIT) is loaded from a CDN, so an internet connection is needed when the file is opened. The Egas Moniz logotype is excluded from both licences. Version 1.1, 2 October 2026.SHA-256: 49ad91bbfde22040efd28e8bf8e52068ac6203e566f7cf06d83af1dbfa232ab0

This is a new feature release, which adds detector duty cycles and fixes a bug with PSDs for O4. It is not believed to introduce any breaking changes. Changes Detector duty cycles A new minke.duty_cyc…

This is a new feature release, which adds detector duty cycles and fixes a bug with PSDs for O4. It is not believed to introduce any breaking changes. Changes Detector duty cycles A new minke.duty_cycle module models each detector as a two-state (locked/unlocked) renewal process with exponentially-distributed segment durations. generate_duty_cycle_schedule draws a schedule over an observation window for a target duty cycle and mean lock duration, and active_detectors restricts a detector network to those locked at a given GPS time, so that which detectors observe an injection varies realistically (and with time-correlation) across a run. Bug-fixes O4 PSDs "AdvancedLIGO_O4" previously resolved to lalsimulation.SimNoisePSDaLIGOAdVO4T1800545, which is a deprecated alias for Virgo's O4 curve, so LIGO detectors were silently given Virgo's sensitivity. It now uses the aLIGO design-sensitivity curve, and a new "AdvancedVirgo_O4" PSD provides Virgo's O4 curve explicitly. Breaking changes This release is not believed to introduce any breaking changes.