Les capitalistes sociaux sur Twitter : détection, évolution, caractérisation

Nicolas Dugué

Jeudi 07 novembre 2013 Ă  11h, salle 25-26/101

Slides

Les capitalistes sociaux sont des utilisateurs particuliers de Twitter. Ces utilisateurs cherchent Ă  obtenir un maximum de followers par des mĂ©thodes que nous dĂ©crirons pour gagner de la visibilitĂ© sur ce rĂ©seau. La visibilitĂ© et la potentielle influence obtenues par ces utilisateurs ne sont pas basĂ©es sur le contenu de leurs tweets et la crĂ©dibilitĂ© de leur compte mais sur une accumulation de followers artificielle. Il est donc intĂ©ressant de dĂ©tecter ces utilisateurs afin d’Ă©tudier leur rĂ©elle influence sur le rĂ©seau. Nous proposons une mĂ©thode de dĂ©tection des capitalistes sociaux utilisant des mesures simples basĂ©es sur la topologie du rĂ©seau uniquement. Suite Ă  cela, nous montrons que les mĂ©thodes employĂ©es par ces utilisateurs font qu’ils forment un sous groupe densĂ©ment connectĂ© dans le graphe reprĂ©sentant le rĂ©seau. Par ailleurs, Ă  travers une Ă©tude sur l’Ă©volution de certains de ces comptes entre 2009 et 2013, nous dĂ©montrons l’efficacitĂ© de ces techniques pour accumuler des followers. Nous confirmons ensuite grâce Ă  un compte Twitter automatisĂ© qu’il est toujours possible d’appliquer ces mĂ©thodes. Enfin, nous nous intĂ©ressons Ă  la position des capitalistes sociaux dans le rĂ©seau. Nous nous basons ainsi sur la notion de rĂ´les communautaires introduite par GuimerĂ  et Amaral pour caractĂ©riser la position de ces utilisateurs au sein des communautĂ©s du rĂ©seau. Nous gĂ©nĂ©ralisons cette mĂ©thode, l’adaptons aux graphes orientĂ©s et montrons que les capitalistes sociaux occupent des rĂ´les spĂ©cifiques.

Analyse des rĂ©seaux et gĂ©ographie politique : l’ONU comme terrain de jeu

Laurent Beauguitte

Jeudi 24 octobre 2013 Ă  11h, salle 26-00/428

Slides

Si la gĂ©ographie a longtemps et de manière quasi exclusive privilĂ©giĂ© l’Ă©tude des rĂ©seaux techniques (rĂ©seaux de transport notamment, voir Barthelemy, 2011), l’analyse de rĂ©seau, entendue comme une boĂ®te Ă  outils mĂ©thodologiques plus que comme une thĂ©orie des phĂ©nomènes sociaux, permet d’enrichir les approches en gĂ©ographie Ă©conomique ou politique. Cette prĂ©sentation montre comment divers outils et mesures, issus de la Social network analysis comme des Complex network studies, peuvent ĂŞtre mobilisĂ©s pour une rĂ©flexion relative Ă  la rĂ©gionalisation politique du monde. La première partie prĂ©sente le cadre Ă©pistĂ©mologique et les emprunts disciplinaires effectuĂ©s. Diverses hypothèses relatives Ă  la rĂ©gionalisation politique sont ensuite exposĂ©es ainsi que les principaux rĂ©sultats obtenus avec des donnĂ©es issues de l’AssemblĂ©e gĂ©nĂ©rale de l’ONU (vote et parrainage de rĂ©solution, lien entre tats et groupes rĂ©gionaux). Enfin, une troisième partie souligne les limites conceptuelles et mĂ©thodologiques des choix effectuĂ©s et de possibles pistes de recherche permettant de les contourner. Si la pluri-disciplinaritĂ© paraĂ®t une voie prometteuse, les obstacles demeurent et ne se limitent pas Ă  des choix lexicaux divergents. RĂ©fĂ©rences Marc Barthelemy, 2011, Spatial networks, Physics reports, 499, 1-101. Laurent Beauguitte, 2011, L’AssemblĂ©e gĂ©nĂ©rale de l’ONU de 1985 Ă  nos jours. Essai de gĂ©ographie politique quantitative, Thèse de doctorat, UniversitĂ© Denis Diderot Paris 7, disponible sur TEL.

The Random Subgraph Model for the analysis of an ecclesiastical network in Merovingian Gaul

Charles Bouveyron

Jeudi 03 octobre 2013 Ă  11h, salle 25-26/101

Slides

In the last two decades, many random graph models have been proposed to extract knowledge from networks. Most of them look for communities or more generally clusters of vertices with homogeneous connection profiles. While the first models focused on networks with binary edges only, extensions now allow to deal with valued networks. Recently, new models were also introduced in order to characterize connection patterns in networks through mixed memberships. This work was motivated by the need of analyzing a historical network where a partition of the vertices is given and where edges are typed. A known partition is seen as a decomposition of a network into subgraphs that we propose to model using a stochastic model with unknown latent clusters. Each subgraph has its own mixing vector and sees its vertices associated to the clusters. The vertices then connect with a probability depending on the subgraphs only, while the types of the edges are assumed to be sampled from the latent clusters. A variational Bayes expectation-maximization algorithm is proposed for inference as well as a model selection criterion for the estimation of the cluster number. Experiments are carried out on simulated data to assess the approach. The proposed methodology is then applied to an ecclesiastical network in merovingian Gaul. An R package, called Rambo, implementing the inference algorithm is available on the CRAN. This is a joint work with Y. Jernite, P. Latouche, P. Rivera, L. Jegou & S. Lamassé. Preprint available at http://arxiv.org/abs/1212.5497.

The Power of Consensus: Random Graphs Have No Communities

Romain Campigotto, Jean-Loup Guillaume and Massoud Seifi

Proceedings of the 5th IEEE/ACM International Conference on Advances in Social Networks and Mining (ASONAM 2013). Niagara Falls, Canada.

Communities are a powerful tool to describe the structure of complex networks. Algorithms aiming at maximizing a quality function called modularity have been shown to effectively compute the community structure. However, some problems remain: in particular, it is possible to find high modularity partitions in graph without any community structure, in particular random graphs. In this paper, we study the notion of consensual communities and show that they do not exist in random graphs. For that, we exhibit a phase transition based on the strength of consensus: below a given threshold, all the nodes belongs to the same consensual community; above this threshold, each node is in its own consensual community.

Download

Assessing Group Cohesion in Homophily Networks

Benjamin Renoust

Mardi 17 septembre 2013 Ă  11h, salle 25-26/101

Slides

The analysis and exploration of a social network depends on the type of relations at play. Borgatti had proposed a type taxonomy organizing relations in four possible categories. Homophily (similarity) relationships form an important category where relations occur when entities of the network link whenever they exhibit similar behaviors. Examples are networks of co-author, where homophily between two persons follows from co-authorship; or network of actors having played under the supervision of the same movie director, for instance. Homophily is often embodied through a bipartite network where entities of a given type A (authors, movie directors) connect through entities of a different type B (papers, actors). A common strategy is then to project this bipartite graph onto a single-type network with entities of a same type A , possibly weighting edges based on how the type A entities interact with the type B entities underlying the edge. The resulting single-type network can then be studied using standard techniques such as community detection using edge density, or the computation of various centrality indices. This paper revisits this type of approach and introduces three measures derived from past work by Burt. Two entities of type B interact when they both induce a same edge between two entities of type A . The homogeneity of a subgroup thus depends on how intensely and how equally interactions occur between entities of type B giving rise to the subgroup. The measure thus differentiates between subgroups of type A exhibiting similar topologies depending on the interaction patterns of the underlying entities of type B.

Towards realistic modeling of IP-level routing topology dynamics

Clémence Magnien, Amélie Medem, Sergey Kirgizov, Fabien Tarissan

Networking Science, 4 (1-4), p. 24-33, 2013

Many works have studied the Internet topology, but few have investigated the question of how it evolves over time. This paper focuses on the Internet routing IP-level topology and proposes a first step towards realistic modeling of its dynamics. We study periodic measurements of routing trees from a single monitor to a fixed destination set and identify invariant properties of its dynamics. Based on those observations, we then propose a model for the underlying mechanisms of the topology dynamics. Our model remains simple as it only incorporates load-balancing phenomena and routing changes. By extensive simulations,  we show that, despite its simplicity, this model effectively captures the observed behaviors, thus providing key insights of relevant mechanisms governing the Internet routing dynamics. Besides, by confronting simulations over different kinds of topology, we also provide insights of which structural properties play a key role to explain the properties of the observed dynamics, which therefore strengthens the relevance of our model.

Download

The network of the International Criminal Court decisions as a complex system

Fabien Tarissan and Raphaëlle Nollez-Goldbach

ISCS 2013: Interdisciplinary Symposium on Complex Systems, Emergence, Complexity and Computation, 8:225-264, Springer, 2013.

Many real-world networks lend themselves to the use of graphs for analysing and modeling their structure. This approach has proved to be very useful for a wide variety of networks stemming from very different fields. Yet, only few papers focused their attention on legal networks. This paper intends precisely to remedy this situation by analysing a major legal network by means of complex system methods. The network under investigation is the network composed by decisions taken by the International Criminal Court since its creation. We first model the network by a simple directed graph in which nodes are the decisions and links represent citations between decisions. Our analysis shows that standard properties shared by common real networks are also present in this network. Then we turn to studying the network by means of bipartite graphs that involve both decisions and articles of law. We show that this two-level structure presents several non trivial properties and we show evidences of the relevance of the bipartite representation to explain properties observed in the graph of citations.

Download

A Matter of Time – Intrinsic or Extrinsic – for Diffusion in Evolving Complex Networks

Alice Albano, Jean-Loup Guillaume, Sébastien Heymann and Bénédicte Le Grand

Proceedings of the 2013 IEEE/ACM International Conference on Advances  n Social Networks Analysis and Mining (ASONAM 2013), Niagara Falls, Canada

Diffusion phenomena occur in many kinds of real-world complex networks, e.g., biological, information or social networks. Because of this diversity, several types of diffusion models have been proposed in the literature: epidemiological models, threshold models, innovation adoption models, among others. Many studies aim at investigating diffusion as an evolving phenomenon but mostly occurring on static networks, and much remains to be done to understand diffusion on evolving networks. In order to study the impact of graph dynamics on diffusion, we propose in this paper an innovative approach based on a notion of intrinsic time, where the time unit corresponds to the appearance of a new link in the graph. This original notion of time allows us to isolate somehow the diffusion phenomenon from the evolution of the network. The objective is to compare the diffusion features observed with this intrinsic time concept from those obtained with traditional (extrinsic) time, based on seconds. The comparison of these time concepts is easily understandable yet completely new in the study of diffusion phenomena. We experiment our approach on synthetic graphs, as well as on a dataset extracted from the Github sofware sharing platform.

Download

valuation et optimisation d’une partition hiĂ©rarchique de graphe

François Queyroi

Mardi 09 juillet 2013 Ă  14h, salle 25-26/101

Slides

Des travaux en sociologie, gĂ©ographie ou biologie suggèrent la prĂ©sence d’une structure de communautĂ©s multi-niveaux au sein des rĂ©seaux complexes. Cette structure peut ĂŞtre modĂ©lisĂ©e par un partitionnement hiĂ©rarchique des sommets d’un graphe. Plusieurs algorithmes ont Ă©tĂ© proposĂ©s rĂ©cemment pour rĂ©pondre Ă  ce problème. En revanche, la question de l’Ă©valuation d’une partition hiĂ©rarchique a Ă©tĂ© peu Ă©tudiĂ©e. Je prĂ©senterai une gĂ©nĂ©ralisation des mesures de qualitĂ© additives au partitionnements multi-niveaux. Cette gĂ©nĂ©ralisation sinterprète comme un parcours des nĹ“uds de l’arbre de partition rĂ©alisĂ© en propageant le « gain » de chaque groupe Ă  ses descendants. Je discuterai Ă©galement plusieurs applications possible utilisant ce nouveau type de mesure ; notamment l’optimisation de la hiĂ©rarchie produite lors du dĂ©roulement de l’algorithme de Louvain.

Using the Framework of Networks to Enhance Learning and Social Interactions

Dmitry Paranyushkin

Jeudi 27 juin 2013 Ă  11h, salle 55-65/211

Slides

The increasingly interconnected world brings up the new challenges related to rapid defragmentation of information and cognitive overload. The existing recommender systems and social networks tend to pack concepts and people into tightly-knit interest communities producing so-called filter bubbles » (Pariser 2011), making it difficult for such systems to evolve, adapt, and innovate. To address those challenges, we developed several social interaction strategies and online tools that are aimed at creating the new possibilities for communication and learning. The intention is to find out how the framework of networks can be used to enhance our learning strategies and expand ones capabilities for social interactions. Specifically, were interested in the notion of metastability the ability of a dynamical system to maintain several distinct latent states at once, which can interact and produce complex behavior on the global level. Metastable dynamics has been shown to be essential to adaptability of a complex system, which has to respond to the constantly changing environment. In this seminar we will present several case studies conducted by Nodus Labs. One of the projects we will present to exemplify our ideas is the online text network visualization tool – http://textexture.com – which can be used to represent any text as a network of interrelated concepts. The graph can then be used to get a general idea or a summary of the texts content, as well as the relations between the different topics present within the text. It can also be used for non-linear fast reading, allowing the users to create different narratives that are more relevant to their fields of interest. We will also present several case studies from our workshop and educational practice (see http://noduslabs.com for more information), where we created so-called constructed situations. In those carefully designed social settings we invited the participants to explore the basic ideas of network dynamics and metastability. The intention was to demonstrate how network thinking can be used to increase ones choices in any social or collaborative situation and lead to a better awareness of communicative dynamics within a group of people.

Scalable Analysis for Network Monitoring and Forensics Purposes

Jérôme François

Jeudi 06 juin 2013 Ă  11h30, salle 55-65/211

Slides

Security issues in Internet force the deployment of defensive measures to protect end users and Internet’s infrastructure itself. While a simple firewall would have been enough in the past, the trend is to promote a deeper analysis nowadays, in particular at the Internet operator level. Simple filtering has to be completed using more in-depth analysis tool. Detection of attacks may have to investigate multiple sources of data meantime and such sources, like network traffic captures, syslog, alerts or locations, may generate huge quantities of data. Forensics alleviates the real-time constraint but requires a perfect and global understanding of an intrusion to recover, protect in future and trigger legal actions as well. Hence, the problem is similar and finding evidences is like looking for a needle in a haystack. Therefore, the seminar will introduce several techniques to cope with big data issues in the context of security. Firstly, flow based methods will be presented as, for example, to track community of hosts participating to a botnet. This is possible by analyzing the traffic flow dependency in Internet and host relationships. Cyber-criminal organizations, like the Russian Business Network, are well organized and constructs their own Internet infrastructure and administrative domains which make them quite resistant to standard counter-measures like IP blacklisting. The seminar will then highlight how to reveal the underlying organization structure at a the Internet administrative domain level.

Towards a Bipartite Graph Modeling of the Internet Topology

Fabien Tarissan, Bruno Quoitin, Pascal Mérindol, Benoit Donnet, Matthieu Latapy et Jean-Jacques Pansiot

In Journal of Computer Networks, 57(11):2331-2347, Elsevier, 2013.

Modeling the properties of the Internet topology aims at generating large scale artificial IP networks that mimic properties of real ones for simulation purposes. Current models typ- ically consider the Internet as a simple graph where edges are point-to-point connections between routers. This approach does not take into account point-to-multipoint connec- tions that exist at lower layers in the network, e.g. layer-2 clouds, such as Ethernet switches or MPLS networks. Instead, such physical point-to-multipoint connections are modeled as several logical IP level point-to-point connections. In this paper, we rely on recent developments in topology discovery based on IGMP probing that allows for revealing part of the network’s layer-2 structure. We take advantage of this additional knowledge for proposing an Internet model based on bipartite graphs considering both point-to-point and point-to-multipoint connections. Our model remains simple: it only takes as input the node degree sequence for both layer-2 and layer-3 nodes, randomly generates a bipartite graph respecting those distri- butions, and then derives the corresponding layer-3 topology. We show that, despite the simplicity of our model, realistic network properties, such as high local density, emerge naturally. This is in contrast with the now common belief that such properties can only appear with more intricate models or if explicitly injected in random models. Besides, we also provide evidences of how the analysis performed at the bipartite level might shed light on important properties of the real network structure. Finally, we propose and evaluate a bipartite graph generator based on our model that only takes two synthetic node degree distributions as input.

Download

Social Network Analysis of Authority in the Blogosphere and its Application

Darko Obradovic

Lundi 3 juin 2013 Ă  11h, salle 25-26/101

Blogs are among the first social media sources in the Web 2.0, and they remain influential until today, with a broad coverage of topics and languages. Due to their decentralised structure, sampling of data and network analyses are different from online social networking sites. We present a possible method and evaluation for identifying and measuring authoritative blogs with SNA, using k-cores, random graphs and community identification. These results are then applied in a prototype tool for the monitoring of specific topics, in combination with text-based subtopic detection, polarity classification and a trend detection.

Partition en sous-graphes denses pour la détection de communautés

Julien Darlay

Jeudi 23 mai 2013 Ă  11h, salle 25-26/101

Slides

La dĂ©tection de communautĂ©s est un problème d’analyse de donnĂ©es oĂą les informations peuvent ĂŞtre reprĂ©sentĂ©es comme un graphe. Les sommets correspondent aux observations et les arĂŞtes reprĂ©sentent des interactions entre les observations. On cherche gĂ©nĂ©ralement une partition des sommets du graphe en classes induisant des sous-graphes denses, c’est-Ă -dire des groupes d’observations presque toutes deux Ă  deux similaires. Dans ce contexte, nous proposons une fonction objectif pour le problème de partition de graphe basĂ©e sur la densitĂ© dĂ©finie par Goldberg. La densitĂ© d’un graphe est le rapport entre le nombre d’arĂŞtes et le nombre de sommets. La densitĂ© d’une partition d’un graphe est alors dĂ©finie comme la somme des densitĂ©s des sous-graphes induits par chaque classe de la partition. Nous montrons que le problème consistant Ă  trouver la partition de densitĂ© maximale est un problème NP-difficile et non approximable. Lorsque le graphe est un arbre, nous montrons qu’il existe un algorithme polynomial pour trouver la partition optimale. Nous proposons une heuristique Ă  base de recherche locale Ă  l’aide de LocalSolver que nous Ă©valuons sur des instances de la littĂ©rature.

Unfolding ego-centered community structures with « a similarity approach »

Maximilien Danisch, Jean-Loup Guillaume and Bénédicte Le Grand

CompleNet 2013, Berlin

We propose a framework to unfold the ego-centered community structure of a given node in a network. The framework is not based on the optimization of a quality function, but on the study of the irregularity of the decrease of a similarity measure. It is a practical use of the notion of multi-ego-centered community and we validate the pertinence of the approach on a real-world network of wikipedia pages.

Download

Internet routing paths stability model and relation to forwarding paths

Dimitri Papadimitriou, Davide Careglio, Fabien Tarissan and Piet Demeester

Proceedings of the the 9th International Conference on Design of Reliable Communication Networks (DRCN), Budapest, Hungary, 2013

Analysis of real datasets to characterize the local stability properties of the Internet routing paths suggests that extending the route selection criteria to account for such property would not increase the routing path length. Nevertheless, even if selecting a more stable routing path could be considered as valuable from a routing perspective, it does not necessarily imply that the associated forwarding path would be more stable. Hence, if the dynamics of the Internet routing and forwarding system show different properties, then one can not straightforwardly derive the one from the other. If this assumption is verified, then the relationship between the stability of the forwarding path (followed by the traffic) and the corresponding routing path as selected by the path-vector routing algorithm requires further characterization. For this purpose, we locally relate, i.e., at the router level, the stability properties of routing path with the corresponding forwarding path. The proposed stability model and measurement results verify this assumption and show that, although the main cause of instability results from the forwarding plane, a second order effect relates forwarding and routing path instability events. This observation provides the first indication that differential stability can safely be taken into account as part of the route selection process.

Download

Wavelets on Graphs: a Tool for Multiscale Community Mining in Graphs

Nicolas Tremblay

Jeudi 11 avril 2013 Ă  11h, salle 25-26/101

Slides

For data represented by networks, the community structure of the underlying graph is of great interest. A classical clustering problem is to uncover the overall best partition of nodes in communities. We work on a more elaborate description in which community structures are identified at different scales. To this end, we take advantage of the local and scale-dependent information encoded in graph wavelets. We classify nodes according to their wavelets or scaling functions, using, for instance, a scale-dependent modularity function. I will give an introduction on spectral graph wavelets and scaling functions, and talk about our recent advances. I will show results obtained on a graph benchmark having hierarchical structure and on real social networks. This is joint work with my supervisor Pierre Borgnat.

Propriétés combinatoires et de robustesse de modèles discrets de réseaux biologiques

Sylvain Sené

Mercredi 13 mars 2013 Ă  14h, salle 25-26/105

Slides

Les rĂ©seaux d’automates sont des objets mathĂ©matiques mettant en jeu des entitĂ©s (dites automates) qui interagissent les unes avec les autres au cours d’un temps discret. En voyant ces rĂ©seaux comme des modèles potentiels de systèmes d’interactions biologiques, l’idĂ©e gĂ©nĂ©rale de cet exposĂ© est de montrer que l’informatique fondamentale permet d’accroĂ®tre la connaissance des lois gĂ©nĂ©rales qui rĂ©gissent le vivant. Plus prĂ©cisĂ©ment, nous utiliserons les rĂ©seaux d’automates boolĂ©ens comme modèles de rĂ©seaux de rĂ©gulation gĂ©nĂ©tique. Dans ce cadre, nous focaliserons notre attention sur deux thèmes, dĂ©veloppĂ©s en collaboration avec Mathilde Noual (I3S, UNS) et Damien Regnault (IBISC, UEVE) : – la combinatoire comportementale des cycles, objets dont on connaĂ®t l’importance sur la dynamique des rĂ©seaux depuis les travaux de RenĂ© Thomas (1981) et de François Robert (1986), et – la robustesse structurelle des rĂ©seaux, au sens de RenĂ© Thom (1972), que nous aborderons au travers de l’influence des modes de mise Ă  jour, et qui nous mènera Ă  l’Ă©tude d’une famille particulière de rĂ©seaux, les rĂ©seaux xor circulants.

Lois d’Ă©chelle des processus de trafic dans les rĂ©seaux de communications

Paulo Gonçalves

Jeudi 07 mars 2013 Ă  14h30, salle 25-26/101

Slides

Les travaux pionniers de Paxson (1994) et de Leland (1994), ont mis en Ă©vidence l’existence et identifiĂ© l’origine physique des propriĂ©tĂ©s d’auto-similaritĂ© et de dĂ©pendance Ă  longue portĂ©e dans les signaux de trafic agrĂ©gĂ©. Mais ces comportements ne sont pas les seules manifestations de phĂ©nomènes d’invariance d’Ă©chelle que l’on peut observer dans les rĂ©seaux de communications. Notamment, nous montrerons que le trafic agrĂ©gĂ© prĂ©sente en fait deux rĂ©gimes de dĂ©pendance Ă  long terme, d’origines diffĂ©rentes et correspondant chacun Ă  une gamme d’Ă©chelle d’agrĂ©gation propre. Nous nous intĂ©resserons ensuite Ă  un flot TCP individuel et montrerons que celui-ci vĂ©rifie un principe empirique de grandes dĂ©viations que l’on sait caractĂ©riser analytiquement via un modèle de Markov. Ce rĂ©sultat nous permet en particulier de gĂ©nĂ©raliser la relation dite de Padhye Ă  une distribution arbitraire des pertes de paquets. Dans un autre registre, nous proposerons enfin un modèle permettant de simuler la volatilitĂ© de charge d’un serveur de VidĂ©os Ă  la Demande, mais qui vĂ©rifie un principe analogue de grandes dĂ©viations. Pour finir, nous ouvrirons alors quelques pistes de rĂ©flexion sur l’exploitation de ces propriĂ©tĂ©s d’invariance d’Ă©chelle particulières pour dĂ©finir des politiques de management probabiliste des ressources. Travaux menĂ©s en collaboration avec P. Loiseau, S. Roy, T. Begin et J. Barral.

Connectivity of Bluetooth Graphs

Nicolas Broutin

Jeudi 28 mars 2013 Ă  11h, salle 25-26/101

Slides

One of the main models for wireless networks is the random geometric graph. In this model, the graph gets connected with high probability only when the average degree is of the order of the logarithm of the size. Although it is not enourmous, it still raises the question of the scalability. Other models (irrigation graphs or Bluetooth graphs) have been devised that sparsify the graph using a local rule and hope that it remains connected. We prove tight threshold for the number of edges necessary for connectivity in this model, showing that the average degree must in particular tend to infinity to expect connectivity. This is joint work with L. Devroye, N. Fraiman and G. Lugosi.