BEGIN:VCALENDAR VERSION:2.0 PRODID:-//Pentabarf//Schedule 0.3//EN CALSCALE:GREGORIAN METHOD:PUBLISH X-WR-CALDESC;VALUE=TEXT:Graph Processing devroom X-WR-CALNAME;VALUE=TEXT:Graph Processing devroom X-WR-TIMEZONE;VALUE=TEXT:Europe/Brussels BEGIN:VEVENT METHOD:PUBLISH UID:8648@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T103000 DTEND:20190202T111000 SUMMARY:AMENDMENT Introduction of OSS Weaviate, the Decentralised Knowledge Graph DESCRIPTION:
Weaviate is an open source decentralized knowledge graph. During this talk, I will introduce the software Weaviate, present specific use cases, present Weaviate's architecture, and introduce one of the core features: the contextionary. More info about Weaviate: https://github.com/creativesoftwarefdn/weaviate
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_weaviate_knowledge_graph/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Etienne Dilocker":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:8658@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T111500 DTEND:20190202T115500 SUMMARY:Gephi JS: Exploring the dystopian future of a Javascript Gephi DESCRIPTION:Gephi is a popular network analysis open source software written in Java. This talk about its future is a joint effort by the Gephi dev team and Javascript experts to explore the issues and opportunities of web technologies for large-scale network analysis and visualization.
Gephi relies on the Swing library for the general user interface plus an OpenGL insert for network visualization. This choice made more than 10 years ago was one of the very few options available for a multiplatform software able to harness GPU power for displaying millions of nodes and edges. Despite a number of issues, the situation was considered relatively satisfying.
However two long-term trends force the Gephi dev team to reconsider its future: computer infrastructures are moving away from true multiplatform software to web-based interfaces, and Java is moving away from user interface development to server-side computation. Most if not all UX developers agree that web technologies are better supporting modern UX design than the Java ecosystem. However the specific use of OpenGL puts Gephi is a special situation. We know from user surveys that scaling up your network is key to most users. No Gephi user has only small networks, and many of them has tried to analyze networks with hundred thousands nodes or more. This use case is far from usual in the world of web technologies. Is WebGL up to the task of visualizing such amount of objects?
We used existing Javascript libraries to build a stack comparable to the basic features of Gephi. Graphology was used as a Javascript equivalent of Gephi's GraphStore library, and a tweaked version of Sigma JS WebGL renderer was used as display engine. We used this prototype to benchmark the performance of web technologies for rendering large graphs in OpenGL.
In parallel we explored the possibility to keep the project in the Java world but stop relying on the JOGL library that is at risk of being less and less maintained. In this mind we developed a new OpenGL visualisation engine prototype, with a better code base and better performance than current implementation.
In this talk we will showcase both the Javascript proof-of-concept and the new Java prototype. We will present state of our reflections and the results of our benchmarks. Of course technological choices are not only (and even not primarily) a question of performance, and we will take the time to discuss the various issues and limits we met in this exercise of prospective technological exploration. we will conclude with a broader reflection on how Gephi as an open source project tries to adapt to global technological changes.
We expect this talk to be useful to Java and/or Javascript developers/engineers dealing with graphs.
Eduardo Ramos Ibáñez is lead developer of GephiMathieu Jacomy is co-author of GephiAlexis Jacomy is co-author of the Sigma JS libraryGuillaume Plique is author of Graphology and co-author of Sigma JS
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_gephi_js/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Mathieu Jacomy":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:8646@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T120000 DTEND:20190202T124000 SUMMARY:Leveraging real-time streaming with Neo4j-Streams DESCRIPTION:Nowadays real-time Messaging Systems has become widely used, covering a variety of use-cases, from log aggregation to real-time Stream Processing (in combination with Big Data Processing Frameworks).In this talk we'll introduce how to use Apache Kafka (the most used Message Brocker) in combination with Neo4j through the Neo4j-Streams project, demonstrating via simple use-cases how you can leverage the information driven by the Change Data Capture Module and how to add Neo4j in your streaming flow by using the Sink module.
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_real_time_streaming_kafka_neo4j/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Andrea Santurbano":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:8628@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T124500 DTEND:20190202T130500 SUMMARY:Graph usage in EFL DESCRIPTION:EFL is a graphical framework that is used on a lot of TVs. The challenge with a TV is, that it needs to be operated with a remote control. Which means, every selectable element on the screen has to be connected to a parent element (i.e. it needs to have a way to navigate to and from itself). In other words, it can be accessed just by pressing the 4 navigation keys on your remote control.
To achieve this, graphical elements on the screen are grouped into graphs, which are then connected together after the creation process.
The talk will cover:
· how the set of graphical components (or elements) on the screen are separated into smaller graphs;· look at how they are connected together; and· show why caching calculation results is now easier to cache, than in the previous system.
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_usage_efl/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Marcel Hollerbach":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:8561@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T131000 DTEND:20190202T135000 SUMMARY:Using networks to study 18th century French trade DESCRIPTION:France started to compile statistics about its trade in 1716.The "Bureau de la Balance du Commerce" (Balance of Trade's Office) centralized local reports of imports/exports by commodities produced by french tax regions.Many statistical manuscript volumes produced by this process have been preserved in French archives.This communication will relate how and why we used network technologies to create a research instrument based on the transcriptions of those archives in the TOFLIT18 research project.Our corpus composed of more than 500k yearly trade transactions of one commodity between a French local tax region or a foreign country between 1718 and 1838.We used a graph database to modelize it as a trade network where trade flows are edges between trade partners.We will explain why we had to design a classification system to reduce the heterogeneity of the commodity names and how such a system introduce the need for hyperedges.Our research instruments aiming at providing exploratory data analysis means to researchers, we will present the web application we've built on top of the neo4j database using JavaScript technologies (Decypher, Express, React, Baobab, SigmaJS).We will finally show how graph model was not only a convenient way to store and query our data but also a poweful visual object to explore trade geographical structures and trade products' specialization patterns.
By Paul Girard and Guillaume Plique.Sciences Po, médialab, Paris, France
Project funded by the French Agence Nationale de la Recherche (TOFLIT18)
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_french_trade_study/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Paul Girard":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:7768@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T135500 DTEND:20190202T143500 SUMMARY:Differentiated access control to graph data DESCRIPTION:JanusGraph provides access to persisted graph data in a way that is scalable in data size and graph traversal length. Graphs become more valuable if data from multiple sources is ingested, but in general not all users will be authorized to access all data in the graph, given privacy laws on personal data and corporate policies. Since it is not practical to maintain separate copies of large graphs for the different authorization groups, a technical solution for access control is required. This presentation deals with this issue and discusses techniques to provide access to graph data under heterogeneous confidentiality regimes. In particular, an add-on TinkerPop API is presented that provides an abstration to the required filtering and supports the auditing of gremlin queries for validity.
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_access_control_tinkerpop/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Marc De Lignie":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:7892@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T144000 DTEND:20190202T152000 SUMMARY:Multiplex graph analysis with GraphBLAS DESCRIPTION:Graph analysis workloads present resource-intensive computations that require a large amount of memory and CPU time.Consequently, there an abundance of graph processing tools which build on distributed data processing frameworks, including Spark GraphX, Flink Gelly and Giraph (which runs on Hadoop).According to a recent survey, most of these systems build on the vertex-centric programming model, originally introduced in Google’s Pregel paper.This model defines graph analytical algorithms in terms of vertices communicating with their neighbours through message passing, which allows both easy parallelization (for the systems) and intuitive formalization of the computation (for developers). While these systems indeed exhibit horizontal scalability, they introduce numerous inefficiencies requiring a large amount of resources even for moderately sized graphs.Most practical applications only use graphs up to a few hundred million vertices and edges, which can now be stored comfortably on a single machine. For such graphs, it is worth investigating techniques that allow their evaluation without the additional cost and complexity of operating a distributed cluster.
The GraphBLAS initiative is an effort to design a set of standard building blocks that allow users to formulate graph computations in the language of linear algebra, using operations on sparse adjacency matrices defined on custom semirings. Since its inception, GraphBLAS has been implemented for multiple languages (e.g. C, C++, and Java).Additionally, GraphBLAS is being designed in collaboration with hardware vendors (such as Intel and Nvidia) to define a standardized set of interfaces, which will allow building specialized hardware components for graph processing in the future.
Graph analysis has a significant overlap with network science, a field that aims to uncover the hidden structural properties of graphs and determine the interplay between their vertices. Most works in network science only study homogeneous (monoplex) graphs, and do not distinguish between different types of vertices and edges. We believe this abstraction is wasteful for most real-life networks, which are heterogeneous (multiplex) and emerge by different types of interactions. To illustrate such analyses, we calculated multiplex clustering metrics on the Paradise papers data set to find interesting entities that were engaged in disproportionately high levels of activities with their interconnected neighbours. We found that even on this relatively small data set (2M vertices and 3M edges), naive implementations did not terminate in days. Hence, we adapted techniques from GraphBLAS to optimize the computations to finish in a few minutes.
This talk gives a brief overview of how linear algebra can be used to define graph computations on monoplex graphs, and how we applied it to speedup the calculation of multiplex graph metrics. We present the lessons learnt while experimenting with sparse matrix libraries inC,Java, andJulia. Our graph analyzer framework is available as open-source.
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_multiplex_analysis_graphblas/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="Gabor Szarnyas":invalid:nomail END:VEVENT BEGIN:VEVENT METHOD:PUBLISH UID:7994@FOSDEM19@fosdem.org TZID:Europe-Brussels DTSTART:20190202T152500 DTEND:20190202T160500 SUMMARY:Mgmt Config: A tale of three graphs DESCRIPTION:Mgmt is a real-time automation tool that is fast and safe. It works by running a DAG (graph) of resources.The DAG is built in real-time, by running a graph-based DSL (language).We'll fill this talk with a number of exciting real-time demos that show how some cleverly applied graph algorithms can solve an incredibly difficult set of problems.We'll finish this off by including a demo of building a real-time, distributed, finite state machine (a distributed graph) using our tool.
CLASS:PUBLIC STATUS:CONFIRMED CATEGORIES:Graph Processing URL:https:/fosdem.org/2019/schedule/2019/schedule/event/graph_mgmt_config/ LOCATION:H.1308 (Rolin) ATTENDEE;ROLE=REQ-PARTICIPANT;CUTYPE=INDIVIDUAL;CN="James Shubin":invalid:nomail END:VEVENT END:VCALENDAR