I am excited to announce that I will be presenting some of my work at the 10th JLESC workshop in Bonn, Germany. My talk will be on “Challenges and opportunities with running AI workloads on HPC systems” where I will present some of my findings on applying large-scale density based clustering with DBSCAN onto HPC systems.
Author: Ernir Erlingsson
New App Published: Schatzküste
We just published another App for iOS and Android which I have been working on in my spare time with the Micheal-Succow Stiftung in Germany. It serves as an audio guide (in German) on a few natural reserves in north-eastern Germany, explaining their Fauna and Flora, also without the requirement that the user be physically present.
F2F in Sofia
Met my colleagues from the DEEP-EST research project in Sofia, Bulgaria. Overall, a very productive meeting on our heterogeneous supercomputer. Also interesting discussions on the recent “Quantum Supremacy” research, as the Quantum Computer is also located at the Jülich Supercomputing Centre, our coordinating partner.
Sofia was fantastic, very friendly people and fantastic local cuisine (blown away to be honest).
Meeting at CERN
A few of my colleagues represent CERN in our research project, applying their research on our co-designed supercomputer prototype. This time our project met at CERN in beautiful Geneva and I had the pleasure of being able to visit the Compact Muon Solenoid (CMS ) 100 meters underground. Usually this is not possible due to radiation, but we had luck as the CMS currently undergoing an update (which will apparently take 2 years). An amazing experience, and one could not help but be in awe of the scale and ambition.
Fog Detection with Deep Learning
Today we started with a Kick-off meeting for a new an exciting Deep Learning research project, a collaborative effort between the University of Iceland, Jülich Supercomputing Centre, and the research arm of the Royal Dutch MET office (KNMI).
For the next months our task will be to train fog classifiers based on millions of images captured by hundreds of cameras along the Dutch highways. Differing lighting conditions, fog-categories, movable cameras, and bird-droppings are among the few things we have to take into consideration for our model.
Face-to-face meeting in Mannheim Germany
Recently the DEEP-EST team met up for a three day meeting in Mannheim Germany where the blueprints for the Extreme-Scale-Booster (ESB) super-computing module where decided (among other things). The ESB module will be one of a kind, and I can´t wait to finally use it for our machine learning applications!
International Geoscience and Remote Sensing Symposium (IGARSS) 2018, Valencia, Spain
I attended the IGARSS conference in Valencia, Spain, where I hosted a High-Performance-Computing (HPC) tutorial with Gabriele Cavallaro from the Jülich Supercomputing Centre. Its a large conference and a professional society of the IEEE.
Additionally, I presented my conference paper which deals with scaling Support Vector Machines (SVMs) for very large datasets using HPC.
Deep Learning Course Supervisor, Jülich Supercomputing Centre, Germany
A two day, hands-on, workshop on Deep Learning (DL), which I co-supervised. Attendees got to know about, and train, Convolutional Neural Networks (CNNs) on the JURECA supercomputer at the Jülich Supercomputing Centre (JSC) using remote sensing input datasets.
For further details I refer to the web page of the course´s main supervisor, Prof. Dr. Morris Riedel.
Echt:Natur – A Mobile App for German Nature Reserves
Am very proud to present Echt:Natur, one of my contractor App projects that has come to fruition. The App was developed for the Michael-Succow-Stiftung to function as a guide for eleven nature reserves distributed over north-eastern Germany. I was mainly responsible for the Android and back-end software development.
Here you can find more information on the mobile application, including download links.
Briar Reaches Version 1.0 Milestone
I am very pleased to announce that Briar´s Android application has finally reached the version 1.0 milestone. Unfortunately I don´t have the time to be involved as I used to be, but I am very happy for my colleagues and the open source community that this was achievable.
Download the app here