Name: The Graph Neural Networks and Graph Application Platforms edition
Start: 2022-04-06T19:00:00+02:00
End: 2022-04-06T22:00:00+02:00
Location: Hyperion Lab

Hi all! Finally, we can host a physical event again after years of waiting.

We can't stay behind with the latest trends in our field; therefore, the topic of this event will be **graphs**. We have two speakers, Miltos Kofinas (UvA) and Ivan Despot (MemGraph) both are experts in their own specific graph subfield; graph neural networks and graph databases. The event is hosted by Hyperion Lab (https://www.hyperionlab.nl/). Hyperion Lab will also sponsor snacks and drinks, and the location has free parking and is close to metro station Bullewijk (IKEA). Pizza is sponsored by MemGraph.

**Bio of Miltos Kofinas:**
Miltos Kofinas is a PhD student in the Video & Image Sense Lab at the University of Amsterdam, supervised by Efstratios Gavves. His research focuses on future spatio-temporal forecasting, with applications on forecasting for autonomous vehicles. His research interests include graph neural networks, temporal dynamics, geometric deep learning, and equivariant representations.
Prior to his PhD, he received a Diploma in Electrical and Computer Engineering from the Aristotle University of Thessaloniki. For his Diploma thesis, he researched the topic of Scene Graph Generation using Graph Neural Networks, supervised by Anastasios Delopoulos and Christos Diou. During his studies, he was a computer vision & machine learning engineer at P.A.N.D.O.R.A. Robotics.

**Abstract of Miltos Kofinas' talk:**
Modelling interactions is critical in learning complex dynamical systems, namely systems of interacting objects with highly non-linear and time-dependent behaviour. A large class of such systems can be formalized as geometric graphs, i.e. graphs with nodes positioned in the Euclidean space given an arbitrarily chosen global coordinate system, for instance, vehicles in a traffic scene. Notwithstanding the arbitrary global coordinate system, the governing dynamics of the respective dynamical systems are invariant to rotations and translations, also known as Galilean invariance. As ignoring these invariances leads to worse generalization, in this work we propose local coordinate systems per node-object to induce roto-translation invariance to the geometric graph of the interacting dynamical system. Further, the local coordinate systems allow for a natural definition of anisotropic filtering in graph neural networks. Experiments in traffic scenes, 3D motion capture, and colliding particles demonstrate the proposed approach comfortably outperforms the recent state-of-the-art.

**Bio of Ivan Despot:**
*Developer Relations Engineer, Memgraph*
Ivan Despot is a Developer Relations Engineer at Memgraph. His passion for mathematics and graph theory inspired him to become part of the Memgraph team and start contributing to the field of graph analytics. Besides graph-based technologies, he is also interested in streaming platforms, stream processing and event-driven development.

Twitter: [https://twitter.com/ivan_g_despot](https://twitter.com/ivan_g_despot)
LinkedIn:[ https://www.linkedin.com/in/ivan-g-despot/](https://www.linkedin.com/in/ivan-g-despot/)
Medium:[ https://gdespot.medium.com/](https://gdespot.medium.com/)

**Abstract of Ivan Despots talk:**
The understanding of complex relationships and interdependencies between different data points is crucial to many decision-making processes.
Graph analytics have found their way into every major industry, from marketing and financial services to transportation. Fraud detection, recommendation engines, and process optimization are some of the use cases where real-time decisions are mission-critical, and the underlying domain can be easily modeled as a graph.
By ingesting data with Apache Kafka and applying graph-based stream processing in real-time, you can perform near-instantaneous graph analytics on vast amounts of data. Graph based stream processing can be used in conjunction with machine learning techniques to perform tasks such as link prediction, node classification and clustering. When it comes to complex networks, it’s often necessary to perform graph algorithms such as calculating the PageRank, identifying communities, traversing relationships, etc. While solutions such as ksqlDB or Apache Spark are useful for processing relational data, Memgraph is an open-source streaming platform that can be used to analyze graph-based data models.
Graph analytics can provide insights into complex networks that would otherwise require resource-intensive computations. It is also much simpler to store streaming data in the form of graphs, as the graph model doesn't rely on predefined and rigid tables. When connecting a Kafka data stream to Memgraph, you only need to create a transformation module that will map incoming messages to the property graph model. This data can then be traversed and analyzed using the Cypher query language without having to implement custom algorithms or relying on development-heavy solutions.

Hope to see you on the 6th!

Gijs Molenaar

Amsterdam Applied Machine Learning

Technology

Artificial Intelligence

Computer Programming

Hadoop

Big Data

Computer Science

Natural Language Processing

Machine Learning

Data Analytics

Data Visualization

Predictive Analytics

Data Mining

Data Science

Artificial Intelligence Applications

Deep Learning

New Technology

The Graph Neural Networks and Graph Application Platforms edition

Graph Databases

Hyperion Lab

Share

Amsterdam Applied Machine Learning

The Graph Neural Networks and Graph Application Platforms edition

Amsterdam Applied Machine Learning

Details

Members are also interested in