Gotland Deep (Baltic Sea) independent age model based on Palaeomagnetic Secular Variation (PSV) and lead (Pb) deposition profiles

Dating of sediment cores from the Baltic Sea has proven to be difficult due to uncertainties surrounding the 14C reservoir age and a scarcity of macrofossils suitable for dating. Here we present the results of multiple dating methods carried out on cores in the Gotland Deep area of the Baltic Sea. Particular emphasis is placed on the Littorina stage (8 ka ago to the present) of the Baltic Sea and possible changes in the 14C reservoir age of our dated samples. Three geochronological methods are used. Firstly, palaeomagnetic secular variations (PSV) are reconstructed, whereby ages are transferred to PSV features through comparison with varved lake sediment based PSV records. Secondly, lead (Pb) content and stable isotope analysis are used to identify past peaks in anthropogenic atmospheric Pb pollution. Lastly, 14C determinations were carried out on benthic foraminifera (Elphidium spec.) samples from the brackish Littorina stage of the Baltic Sea. Determinations carried out on smaller samples (as low as 4 µg C) employed an experimental, state-of-the-art method involving the direct measurement of CO2 from samples by a gas ion source without the need for a graphitisation step - the first time this method has been performed on foraminifera in an applied study. The PSV chronology, based on the uppermost Littorina stage sediments, produced ten age constraints between 6.29 and 1.29 cal ka BP, and the Pb depositional analysis produced two age constraints associated with the Medieval pollution peak. Analysis of PSV data shows that adequate directional data can be derived from both the present Littorina saline phase muds and Baltic Ice Lake stage varved glacial sediments. Ferrimagnetic iron sulphides, most likely authigenic greigite (Fe3S4), present in the intermediate Ancylus Lake freshwater stage sediments acquire a gyroremanent magnetisation during static alternating field (AF) demagnetisation, preventing the identification of a primary natural remanent magnetisation for these sediments. An inferred marine reservoir age offset (deltaR) is calculated by comparing the foraminifera 14C determinations to a PSV & Pb age model. This deltaR is found to trend towards younger values upwards in the core, possibly due to a gradual change in hydrographic conditions brought about by a reduction in marine water exchange from the open sea due to continued isostatic rebound.

Acute survivorship of the deep-sea coral Lophelia pertusa from the Gulf of Mexico under acidification,warming,and deoxygenation

Changing global climate due to anthropogenic emissions of CO2 are driving rapid changes in the physical and chemical environment of the oceans via warming, deoxygenation, and acidification. These changes may threaten the persistence of species and populations across a range of latitudes and depths, including species that support diverse biological communities that in turn provide ecological stability and support commercial interests. Worldwide, but particularly in the North Atlantic and deep Gulf of Mexico, Lophelia pertusa forms expansive reefs that support biological communities whose diversity rivals that of tropical coral reefs. In this study, L. pertusa colonies were collected from the Viosca Knoll region in the Gulf of Mexico (390 to 450 m depth), genotyped using microsatellite markers, and exposed to a series of treatments testing survivorship responses to acidification, warming, and deoxygenation. All coral nubbins survived the acidification scenarios tested, between pH of 7.67 and 7.90 and aragonite saturation states of 0.92 and 1.47. However, calcification generally declined with respect to pH, though a disparate response was evident where select individuals net calcified and others exhibited net dissolution near a saturation state of 1. Warming and deoxygenation both had negative effects on survivorship, with up to 100% mortality observed at temperatures above 14ºC and oxygen concentrations of approximately 1.5 ml·l-1. These results suggest that, over the short-term, climate change and OA may negatively impact L. pertusa in the Gulf of Mexico, though the potential for acclimation and the effects of genetic background should be considered in future research.

Bulk rock magnetic measurements and thermomagnetic analyses of sediment core GeoB4901-8 from the southeastern flank of the Niger deep-sea fan

Diagenesis has extensively affected the magnetic mineral inventory of organic-rich late Quaternary sediments in the Niger deep-sea fan. Changes in concentration, grain size, and coercivity document modifications of the primary magnetic mineral assemblages at two horizons. The first front, the modern iron redox boundary, is characterized by a drastic decline in magnetic mineral content, coarsening of the grain size spectrum, and reduction in coercivity. Beneath a second front, the transition from the suboxic to the sulfidic anoxic domain, a further but less pronounced decrease in concentration and bulk grain size occurs. Finer grains and higher coercive magnetic constituents substantially increase in the anoxic environment. Low- and high-temperature experiments were performed on bulk sediments and on extracts which have also been examined by X-ray diffraction. Thermomagnetic analyses proved ferrimagnetic titanomagnetites of terrigenous provenance as the principal primary magnetic mineral components. Their broad range of titanium contents reflects the volcanogenic traits of the Niger River drainage areas. Diagenetic alteration is not only a grain size selective process but also critically depends on titanomagnetite composition. Low-titanium compounds are less resistant to diagenetic dissolution. Intermediate titanium content titanomagnetite thus persists as the predominant magnetic mineral fraction in the sulfidic anoxic sediments. At the Fe redox boundary, precipitation of authigenic, possibly bacterial, magnetite is documented. The presence of hydrogen sulfide in the pore water suggests a formation of secondary magnetic iron sulfides in the anoxic domain. Grain size-specific data argue for a gradual development of a superparamagnetic and single-domain iron sulfide phase in this milieu, most likely greigite.

Low-temperature magnetic remanence and hysteresis measurements on sediment core GeoB4901-8 from the Niger deep-sea fan

Low-temperature (LT) magnetic remanence and hysteresis measurements, in the range 300-5 K, were combined with energy dispersive spectroscopy (EDS) in order to characterize the magnetic inventory of strongly diagenetically altered sediments originating from the Niger deep-sea fan. We demonstrate the possibility of distinguishing between different compositions of members of the magnetite-ulvöspinel and ilmenite-hematite solid solution series on a set of five representative samples, two from the upper suboxic and three from the lower sulfidic anoxic zone of gravity core GeoB 4901. Highly sensitive LT magnetic measurements were performed on magnetic extracts resulting in large differences in the magnetic behavior between samples from the different layers. This emphasizes that both Fe-Ti oxide phases occur in different proportions in the two geochemical environments. Most prominent are variations in the coercivity sensitive parameter coercive field (BC). At room-temperature (RT) hysteresis loops for all extracts are narrow and yield low coercivities (6-13 mT). With decreasing temperature the loops become more pronounced and wider. At 5 K an approximately 5-fold increase in BC for the suboxic samples contrasts a 20-25-fold increase for the samples from the anoxic zone. We demonstrate that this distinct increase in BC at LT corresponds to the increasing proportion of the Ti-rich hemoilmenite phase, while Fe-rich (titano-)magnetite dominates the magnetic signal at RT. This trend is also seen in the room-temperature saturation isothermal remanent magnetization (RT-SIRM) cycles: suboxic samples show remanence curves dominated by Fe-rich mineral phases while anoxic samples display curves clearly dominated by Ti-rich particles. We show that the EDS intensity ratios of the characteristic Fe Kalpha and Ti Kalpha lines of the Fe-Ti oxides may be used to differentiate between members of the magnetite-ulvöspinel and ilmenite-hematite solid solution series. Furthermore it is possible to calculate an approximate composition for each grain if the intensity ratios of natural particles are linked to well-known standards. Thus, element spectra with high Fe/Ti intensity ratios were found to be rather typical of titanomagnetite while low Fe/Ti ratios are indicative of hemoilmenite. The EDS analyses confirm the LT magnetic results, Fe-rich magnetic phases dominate in the upper suboxic environment whereas Ti-rich magnetic phases comprise the majority of particles in the lower anoxic domain: The mineral assemblage of the upper suboxic environments is composed of magnetite (~19%), titanomagnetite (~62%), hemoilmenite (~17%) and ~2% other particles. In the lower anoxic sediments, reductive diagenetic alteration has resulted in more extensive depletion of the (titano-)magnetite phase, resulting in a relative enrichment of the hemoilmenite phase (~66%). In these strongly anoxic sediments stoichiometric magnetite is barely preserved and only ~5% titanomagnetite was detected. The remaining ~28% comprises Ti-rich particles such as pseudobrookite or rutile.

CO2-driven decrease in pH disrupts olfactory behaviour and increases individual variation in deep-sea hermit crabs

Deep-sea species are generally thought to be less tolerant of environmental variation than shallow-living species due to the relatively stable conditions in deep waters for most parameters (e.g. temperature, salinity, oxygen, and pH). To explore the potential for deep-sea hermit crabs (Pagurus tanneri) to acclimate to future ocean acidification, we compared their olfactory and metabolic performance under ambient (pH 7.6) and expected future (pH 7.1) conditions. After exposure to reduced pH waters, metabolic rates of hermit crabs increased transiently and olfactory behaviour was impaired, including antennular flicking and prey detection. Crabs exposed to low pH treatments exhibited higher individual variation for both the speed of antennular flicking and speed of prey detection, than observed in the control pH treatment, suggesting that phenotypic diversity could promote adaptation to future ocean acidification.

Sr/Ca and Cd/Ca ratios of benthic foraminifera in surface deep-sea sediments

Measurements of Sr/Ca of benthic foraminifera show a linear decrease with water depth which is superimposed upon significant variability identified by analyses of individual foraminifera. New data for Cd/Ca support previous work in defining a contrast between waters shallower and deeper than ~2500 m. Measured element partition coefficients in foraminiferal calcium carbonate relative to sea water (D) have been described by means of a one-box model in which elements are extracted by Rayleigh distillation from a biomineralization reservoir that serves for calcification with a constant fractionation factor (alpha), such that D = (1 - f**alpha)/(l - f), where f is the proportion of Ca remaining after precipitation. A modification to the model recognises differences in element speciation. The model is consistent with differences between D[Sr], D[Ba], and D[Cd] in benthic but not planktonic foraminifera. Depth variations in D for Sr and Ba are consistent with the model, as are differences in depth variation of D[Cd] in calcitic and aragonitic benthic foraminifera. The shallower depth variations may reflect increasing calcification rates with increasing water depth to an optimum of about 2500 m. Observations of unusually lower DCd for some deep waters, not accompanied by similar [Sr], or D[Ba] may be because of dissolution or a calcification response to a lower carbonate saturation state.

Classificazione di Radiografie Toraciche per la Diagnosi del COVID-19 con Reti Convoluzionali e Vision Transformer

Negli ultimi due anni, per via della pandemia generata dal virus Covid19, la vita in ogni angolo del nostro pianeta è drasticamente cambiata. Ad oggi, nel mondo, sono oltre duecentoventi milioni le persone che hanno contratto questo virus e sono quasi cinque milioni le persone decedute. In alcuni periodi si è arrivati ad avere anche un milione di nuovi contagiati al giorno e mediamente, negli ultimi sei mesi, questo dato è stato di più di mezzo milione al giorno. Gli ospedali, soprattutto nei paesi meno sviluppati, hanno subito un grande stress e molte volte hanno avuto una carenza di risorse per fronteggiare questa grave pandemia. Per questo motivo ogni ricerca in questo campo diventa estremamente importante, soprattutto quelle che, con l'ausilio dell'intelligenza artificiale, riescono a dare supporto ai medici. Queste tecnologie una volta sviluppate e approvate possono essere diffuse a costi molto bassi e accessibili a tutti. In questo elaborato sono stati sperimentati e valutati due diversi approcci alla diagnosi del Covid-19 a partire dalle radiografie toraciche dei pazienti: il primo metodo si basa sul transfer learning di una rete convoluzionale inizialmente pensata per la classificazione di immagini. Il secondo approccio utilizza i Vision Transformer (ViT), un'architettura ampiamente diffusa nel campo del Natural Language Processing adattata ai task di Visione Artificiale. La prima soluzione ha ottenuto un’accuratezza di 0.85 mentre la seconda di 0.92, questi risultati, soprattutto il secondo, sono molto incoraggianti soprattutto vista la minima quantità di dati di training necessaria.

Previsione del Successo di Prodotti di Moda Prima della Commercializzazione: un Nuovo Dataset e Modello di Vision-Language Transformer

A differenza di quanto avviene nel commercio tradizionale, in quello online il cliente non ha la possibilità di toccare con mano o provare il prodotto. La decisione di acquisto viene maturata in base ai dati messi a disposizione dal venditore attraverso titolo, descrizioni, immagini e alle recensioni di clienti precedenti. É quindi possibile prevedere quanto un prodotto venderà sulla base di queste informazioni. La maggior parte delle soluzioni attualmente presenti in letteratura effettua previsioni basandosi sulle recensioni, oppure analizzando il linguaggio usato nelle descrizioni per capire come questo influenzi le vendite. Le recensioni, tuttavia, non sono informazioni note ai venditori prima della commercializzazione del prodotto; usando solo dati testuali, inoltre, si tralascia l’influenza delle immagini. L'obiettivo di questa tesi è usare modelli di machine learning per prevedere il successo di vendita di un prodotto a partire dalle informazioni disponibili al venditore prima della commercializzazione. Si fa questo introducendo un modello cross-modale basato su Vision-Language Transformer in grado di effettuare classificazione. Un modello di questo tipo può aiutare i venditori a massimizzare il successo di vendita dei prodotti. A causa della mancanza, in letteratura, di dataset contenenti informazioni relative a prodotti venduti online che includono l’indicazione del successo di vendita, il lavoro svolto comprende la realizzazione di un dataset adatto a testare la soluzione sviluppata. Il dataset contiene un elenco di 78300 prodotti di Moda venduti su Amazon, per ognuno dei quali vengono riportate le principali informazioni messe a disposizione dal venditore e una misura di successo sul mercato. Questa viene ricavata a partire dal gradimento espresso dagli acquirenti e dal posizionamento del prodotto in una graduatoria basata sul numero di esemplari venduti.

End-to-end Deep Metric Learning con Vision-Language Model per il Fashion Image Captioning

L'image captioning è un task di machine learning che consiste nella generazione di una didascalia, o caption, che descriva le caratteristiche di un'immagine data in input. Questo può essere applicato, ad esempio, per descrivere in dettaglio i prodotti in vendita su un sito di e-commerce, migliorando l'accessibilità del sito web e permettendo un acquisto più consapevole ai clienti con difficoltà visive. La generazione di descrizioni accurate per gli articoli di moda online è importante non solo per migliorare le esperienze di acquisto dei clienti, ma anche per aumentare le vendite online. Oltre alla necessità di presentare correttamente gli attributi degli articoli, infatti, descrivere i propri prodotti con il giusto linguaggio può contribuire a catturare l'attenzione dei clienti. In questa tesi, ci poniamo l'obiettivo di sviluppare un sistema in grado di generare una caption che descriva in modo dettagliato l'immagine di un prodotto dell'industria della moda dato in input, sia esso un capo di vestiario o un qualche tipo di accessorio. A questo proposito, negli ultimi anni molti studi hanno proposto soluzioni basate su reti convoluzionali e LSTM. In questo progetto proponiamo invece un'architettura encoder-decoder, che utilizza il modello Vision Transformer per la codifica delle immagini e GPT-2 per la generazione dei testi. Studiamo inoltre come tecniche di deep metric learning applicate in end-to-end durante l'addestramento influenzino le metriche e la qualità delle caption generate dal nostro modello.

A study on tackling visual odometry by a transformer architecture

This dissertation describes a deepening study about Visual Odometry problem tackled with transformer architectures. The existing VO algorithms are based on heavily hand-crafted features and are not able to generalize well to new environments. To train them, we need carefully fine-tune the hyper-parameters and the network architecture. We propose to tackle the VO problem with transformer because it is a general-purpose architecture and because it was designed to transformer sequences of data from a domain to another one, which is the case of the VO problem. Our first goal is to create synthetic dataset using BlenderProc2 framework to mitigate the problem of the dataset scarcity. The second goal is to tackle the VO problem by using different versions of the transformer architecture, which will be pre-trained on the synthetic dataset and fine-tuned on the real dataset, KITTI dataset. Our approach is defined as follows: we use a feature-extractor to extract features embeddings from a sequence of images, then we feed this sequence of embeddings to the transformer architecture, finally, an MLP is used to predict the sequence of camera poses.

Sviluppo di Retrieval-based Chatbot per l'Italiano con Transformer

L’Intelligenza Artificiale negli ultimi anni sta plasmando il futuro dell’umanità in quasi tutti i settori. È già il motore principale di diverse tecnologie emergenti come i big data, la robotica e l’IoT e continuerà ad agire come innovatore tecnologico nel futuro prossimo. Le recenti scoperte e migliorie sia nel campo dell’hardware che in quello matematico hanno migliorato l’efficienza e ridotto i tempi di esecuzione dei software. È in questo contesto che sta evolvendo anche il Natural Language Processing (NLP), un ramo dell’Intelligenza Artificiale che studia il modo in cui fornire ai computer l'abilità di comprendere un testo scritto o parlato allo stesso modo in cui lo farebbe un essere umano. Le ambiguità che distinguono la lingua naturale dalle altre rendono ardui gli studi in questo settore. Molti dei recenti sviluppi algoritmici su NLP si basano su tecnologie inventate decenni fa. La ricerca in questo settore è quindi in continua evoluzione. Questa tesi si pone l'obiettivo di sviluppare la logica di una chatbot help-desk per un'azienda privata. Lo scopo è, sottoposta una domanda da parte di un utente, restituire la risposta associata presente in una collezione domande-risposte. Il problema che questa tesi affronta è sviluppare un modello di NLP in grado di comprendere il significato semantico delle domande in input, poiché esse possono essere formulate in molteplici modi, preservando il contenuto semantico a discapito della sintassi. A causa delle ridotte dimensioni del dataset italiano proprietario su cui testare il modello chatbot, sono state eseguite molteplici sperimentazioni su un ulteriore dataset italiano con task affine. Attraverso diversi approcci di addestramento, tra cui apprendimento metrico, sono state raggiunte alte accuratezze sulle più comuni metriche di valutazione, confermando le capacità del modello proposto e sviluppato.

Deep Water Running para redução da gordura corporal em mulheres na meia idade : estudo de intervenção em Campinas - SP

Universidade Estadual de Campinas . Faculdade de Educação Física

Swine deep bedding ashes as a mineral additive for cement based mortar

The sustainability of intensive swine production demands alternative destinations for the generated residues. Ashes from swine rice husk-based deep bedding were tested as a mineral addition for cement mortars. The ashes were obtained at 400 to 600ºC, ground and sieved through a 325 mesh sieve (# 0.045 mm). The characterization of the ashes included the determination of the index of pozzolanic activity with lime. The ashes were also tested as partial substitutes of Portland cement. The mortars were prepared using a cement:sand proportion of 1:1.5, and with water/cement ratio of 0.4. Three percentages of mass substitution of the cement were tested: 10, 20 and 30%. Mortar performances were assessed at 7 and 28 days determining their compressive strength. The chosen condition for calcinations at the laboratory scale was related to the maximum temperature of 600ºC since the resulting ashes contained vitreous materials and presented satisfactory values for the pozzolanic index under analysis. The pozzolanic activity indicated promising results for ashes produced at 600ºC as a replacement of up to 30% in cement masses.

Modulation of peritoneal macrophage activity by the saturation state of the fatty acid moiety of phosphatidylcholine

To determine the effects of saturated and unsaturated fatty acids in phosphatidylcholine (PC) on macrophage activity, peritoneal lavage cells were cultured in the presence of phosphatidylcholine rich in saturated or unsaturated fatty acids (sat PC and unsat PC, respectively), both used at concentrations of 32 and 64 µM. The treatment of peritoneal macrophages with 64 µM unsat PC increased the production of hydrogen peroxide by 48.3% compared to control (148.3 ± 16.3 vs 100.0 ± 1.8%, N = 15), and both doses of unsat PC increased adhesion capacity by nearly 50%. Moreover, 64 µM unsat PC decreased neutral red uptake by lysosomes by 32.5% compared to the untreated group (67.5 ± 6.8 vs 100.0 ± 5.5%, N = 15), while both 32 and 64 µM unsat PC decreased the production of lipopolysaccharide-elicited nitric oxide by 30.4% (13.5 ± 2.6 vs 19.4 ± 2.5 µM) and 46.4% (10.4 ± 3.1 vs 19.4 ± 2.5 µM), respectively. Unsat PC did not affect anion production in non-stimulated cells or phagocytosis of unopsonized zymosan particles. A different result pattern was obtained for macrophages treated with sat PC. Phorbol 12-miristate 13-acetate-elicited superoxide production and neutral red uptake were decreased by nearly 25% by 32 and 64 µM sat PC, respectively. Sat PC did not affect nitric oxide or hydrogen peroxide production, adhesion capacity or zymosan phagocytosis. Thus, PC modifies macrophage activity, but this effect depends on cell activation state, fatty acid saturation and esterification to PC molecule and PC concentration. Taken together, these results indicate that the fatty acid moiety of PC modulates macrophage activity and, consequently, is likely to affect immune system regulation in vivo.

From Offshore to Onshore: Multiple Origins of Shallow-Water Corals from Deep-Sea Ancestors

Shallow-water tropical reefs and the deep sea represent the two most diverse marine environments. Understanding the origin and diversification of this biodiversity is a major quest in ecology and evolution. The most prominent and well-supported explanation, articulated since the first explorations of the deep sea, holds that benthic marine fauna originated in shallow, onshore environments, and diversified into deeper waters. In contrast, evidence that groups of marine organisms originated in the deep sea is limited, and the possibility that deep-water taxa have contributed to the formation of shallow-water communities remains untested with phylogenetic methods. Here we show that stylasterid corals (Cnidaria: Hydrozoa: Stylasteridae)-the second most diverse group of hard corals-originated and diversified extensively in the deep sea, and subsequently invaded shallow waters. Our phylogenetic results show that deep-water stylasterid corals have invaded the shallow-water tropics three times, with one additional invasion of the shallow-water temperate zone. Our results also show that anti-predatory innovations arose in the deep sea, but were not involved in the shallow-water invasions. These findings are the first robust evidence that an important group of tropical shallow-water marine animals evolved from deep-water ancestors.