Mitteilungen des URZ 4/2002

Informationen des Universitätsrechenzentrums:Zentrale Ausbildungspools und UB-Computerarbeitsplätze: neue Hardware und Linux-Distribution Web-Trust-Center: Single Sign On für Web-Anwendungen Neuer Webmail-Zugang: IMP Version 4 Windows-Paket-Management Neues in der UB Software-News Personalia

No description available.

The original intention of the recommender system is to solve the problem of information explosion, hoping to help users find the content they need more efficiently. In an e-commerce platform, users typically interact with items that they are interested in or need in a variety of ways. For example, buying, browsing details, etc. These interactions are recorded as time-series information. How to use this sequential information to predict user behaviors in the future and give an efficient and effective recommendation is a very important problem. For content providers, such as merchants in e-commerce platforms, more accurate recommendation means higher traffic, CTR (click-through rate), and revenue. Therefore, in the industry, the CTR model for recommendation systems is a research hotspot. However, in the fine ranking stage of the recommendation system, the existing models have some limitations. No researcher has attempted to predict multiple behaviors of one user simultaneously by processing sequential information. We define this problem as the multi-task sequential recommendation problem. In response to this problem, we study the CTR model, sequential recommendation, and multi-task learning. Based on these studies, this paper proposes AMBSN (Attention-based Multi-Behavior Sequential Network). Specifically, we added a transformer layer, the activation unit, and the multi-task tower to the traditional Embedding&MLP (multi-layer perceptron) model. The transformer layer enables our model to efficiently extract sequential behavior information, the activation unit can understand user interests, and the multi-task tower structure makes the model give the prediction of different user behaviors at the same time. We choose user behavior data from Taobao for recommendation published on TianChi as the dataset, and AUC as the evaluation criterion. We compare the performance of AMBSN and some other models on the test set after training. The final results of the experiment show that our model outperforms some existing models. / L’intenzione originale del sistema di raccomandazione è risolvere il problema dell’esplosione delle informazioni, sperando di aiutare gli utenti a trovare il contenuto di cui hanno bisogno in modo più efficiente. In una piattaforma di e-commerce, gli utenti in genere interagiscono con gli articoli a cui sono interessati o di cui hanno bisogno in vari modi. Ad esempio, acquisti, dettagli di navigazione, ecc. Queste interazioni vengono registrate come informazioni di serie temporali. Come utilizzare queste informazioni sequenziali per prevedere i comportamenti degli utenti in futuro e fornire una raccomandazione efficiente ed efficace è un problema molto importante. Per i fornitori di contenuti, come i commercianti nelle piattaforme di e-commerce, una raccomandazione più accurata significa traffico, CTR (percentuale di clic) ed entrate più elevati. Pertanto, nel settore, il modello CTR per i sistemi di raccomandazione è un hotspot di ricerca. Tuttavia, nella fase di classificazione fine del sistema di raccomandazione, i modelli esistenti presentano alcune limitazioni. Nessun ricercatore ha tentato di prevedere più comportamenti di un utente contemporaneamente elaborando informazioni sequenziali. Definiamo questo problema come il problema di raccomandazione sequenziale multi-task. In risposta a questo problema, studiamo il modello CTR, la raccomandazione sequenziale e l’apprendimento multi-task. Sulla base di questi studi, questo documento propone AMBSN (Attention-based Multi-Behavior Sequential Network). In particolare, abbiamo aggiunto uno strato trasformatore, l’unità di attivazione e la torre multi-task al tradizionale modello Embedding&MLP (multi-layer perceptron). Il livello del trasformatore consente al nostro modello di estrarre in modo efficiente le informazioni sul comportamento sequenziale, l’unità di attivazione può comprendere gli interessi degli utenti e la struttura della torre multi-task fa sì che il modello fornisca la previsione di diversi comportamenti degli utenti contemporaneamente. Scegliamo i dati sul comportamento degli utenti da Taobao per la raccomandazione pubblicata su TianChi come set di dati e l’AUC come criterio di valutazione. Confrontiamo le prestazioni di AMBSN e di alcuni altri modelli sul set di test dopo l’allenamento. I risultati finali dell’esperimento mostrano che il nostro modello supera alcuni modelli esistenti.

Die 'Mitteilungen des URZ' enthalten Informationen für die Nutzer des Universitätsrechenzentrums der TU Chemnitz und erscheinen vierteljährlich.:Inhalt Nr. 3/2003: Mitteilungen des URZ - online Windows XP in den URZ-Pools Software unter Windows XP Automatisches Software-Update TUCWiki - TWiki-Einsatz an der TU Chemnitz Hochverfügbare Services mit kimberlite Formularmanagement Ein Windows-Programm auf dem CLiC Kurzinformationen

La toxine bactérienne de Shiga se lie au glycosphingolipide (GSL) globotriaosylcéramide (Gb3) afin d’entrer par endocytose dans les cellules en utilisant une voie dépendante et indépendante de la clathrine. Dans la voie indépendante de la clathrine, la toxine de Shiga réorganise les lipides de la membrane de façon à imposer une contrainte mécanique sur la bicouche, conduisant ainsi à la formation de pic d’invagination d'endocytose profonds et étroits. Mécaniquement ce phénomène n’est pas encore compris, notamment il reste énigmatique, comment se traduisent les propriétés géométriques de l’agrégation des glycosphingolipides GSLS et de la toxine. Dans mon travail de thèse, via l’utilisation de la sous-unité B de la toxine de Shiga (STxB) comme un modèle, différentes espèces moléculaires de son récepteur Gb3 ont été synthétisés avec des structures délibérément choisis. Les études réalisées par imagerie de haute résolution et par la modélisation informatique ont permis d’élucider les contraintes mécano-chimique sous-jacente conduisant à une réorganisation efficace qui a pour résultat l’agrégation de la toxine et la réorganisation des lipides. En combinant des expériences de simulation sur ordinateur de dynamique des particules dissipatives (DPD) et des expériences sur des modèles de membranes cellulaires, nous avons fourni la preuve de l’induction d’une force de fluctuation-membrane, de type « force de Casimir », conduisant à l'agrégation des molécules de toxines associées à la membrane à des échelles de longueur mésoscoiques. Nous avons observé et mesuré, en outre la condensation lipidique induite par la toxine, quantitativement sur des monocouches de Langmuir en utilisant la réflectivité des rayons X (XR) et par la mesure de la diffraction des rayons X par incidence rasante (GIXD), fournissant ainsi une preuve directe de l'hypothèse que la toxine a le potentiel de réduire de façon asymétrique la surface moléculaire sur la partie membranaire exoplasmique, ce qui conduit à une déformation locale de la membrane. Durant ma thèse, nos efforts ont été consacrés à la réalisation de nouveaux glycosphinolipides (GSL) comme outils chimiques à visée biologique. Par ailleurs, une nouvelle stratégie de reconstitution de GSL fonctionnels sur la membrane cellulaire, basée sur une réaction de ligation de type « click » entre un glycosyl-cyclooctyne et un azido-sphingosine a été étudiée. Les résultats obtenus sur les cellules se sont avérés beaucoup moins efficace que ceux in vitro. Une poursuite de l'optimisation de cette méthodologie est actuellement en cours. Une sonde fluorescente du glycosphinolipide Gb3, marquée à l’Alexa Fluor 568 lui-même lié par l'intermédiaire d'un bras PEG-α à la position de la chaîne acyle, a été synthétisée. Cette sonde se lie à la STxB sur couche mince de TLC, mais pas sur des membranes modèles. D'autres améliorations sont discutées. / Bacterial Shiga toxins bind to the glycosphingolipid (GSL) globotriaosylceramide (Gb3) to enter cells by clathrin-dependent and independent endocytosis. In the clathrin-independent pathway, Shiga toxin reorganizes membrane lipids in a way such as to impose mechanical strain onto the bilayer, thus leading to the formation of deep and narrow endocytic pits. Mechanistically how this occurs is not yet understood, and notably how the geometric properties of toxin-GSLs complexes translate into function has remained enigmatic. In my thesis work, using the B-subunit of Shiga toxin (STxB) as a model, different molecular species of its receptor Gb3 have been synthesized with deliberately chosen structures, coupled with high resolution imaging and computational modeling, to understand the underlying mechano-chemical constraints leading to efficient toxin clustering and lipids reorganization. By combining dissipative particle dynamics (DPD) computer simulation and experiments on cell and model membranes, we provided evidence that a membrane fluctuation-induced force, termed Casimir-like force, drives the aggregation of tightly membrane-associated toxin molecules at mesoscopic length scales. Furthermore, toxin-induced lipid condensation was observed and measured quantitatively on Langmuir monolayers using X-ray reflectivity (XR) and grazing incidence x-ray diffraction (GIXD), thereby providing direct evidence for the hypothesis that the toxin has the potential to asymmetrically reduce the molecular area of the exoplasmic membrane leaflet, leading to local membrane deformation. During my PhD, effort was also invested to develop new GSL tools applied to the biological setting. A novel strategy based on the Cu-free click reaction between glycosyl-cyclooctyne and azido-sphingosine was designed with the goal to functionally incorporate GSLs into cellular membranes. Following the synthesis work, click reactions have been performed in solution and on cells. Compared to the former, results on cells were far less efficient. Further optimization is currently ongoing. A fluorescently labeled Gb3 probe with Alexa Fluor 568 coupled via a PEG linker to the α-position of the acyl chain, was synthesized, to which STxB bound on TLCs, but not on model membranes. Further improvements are discussed.