Milner Update

Our system Milner, intended for research in neuroinformatics, is a Cray XC30, based on 2.5 GHz Intel Ivy Bridge 10-core processors and Cray’s Aries interconnect. The aggregate peak performance is 48 TF and the aggregate compute memory is 3.75 TB. The system also includes a Lustre file system with a usable capacity of more than 150 TB.

Sweden's participation in the EU-funded Human Brain Project (HBP) involves research using Milner. Jeanette Hellgren Kotaleski is the deputy leader of the HBP Brain Simulation Platform, which provides software, workflows and models for brain simulation. One of the long-term goals of the HBP is to simulate the entire human brain. As a step towards this goal, accurate models of the different types of neurons in the brain need to be created. Some of the model parameters can be measured directly, but other parameters are as yet unknown. Johannes Hjorth, who is a postdoctoral researcher in Jeanette's research group at the SciLifeLab, based at the Karolinska Institute in Stockholm, is developing and using a workflow where optimization techniques are employed to determine appropriate values for unknown parameters. Johannes says that “The goal of the optimization is to make our models behave similarly to what we observe in experimental data. The parameter space we investigate is so large that we need to use supercomputers for our parameter search. For this project we are using the PDC supercomputer Milner”. Johannes Hjorth uses Milner to determine experimentally unknown parameters through optimization.

PDC also participates as a partner in the Neuromorphic Computing Platform in the Human Brain Project, where the Multisimulation Coordinator (MUSIC) is being developed. MUSIC is a software tool that works as a communication bridge between brain simulation software tools and/or hardware, allowing larger simulations to be built from components. MUSIC has recently been integrated into the Closed Loop Environment (CLE) software from the HBP Neurorobotics Platform.

One of the projects in the Swedish e-Science Research Centre (SeRC) community known as Brain-IT is a “Scalable systems level simulator” – this project is being run by Anders Lansner, Örjan Ekeberg and other researchers. Here the interaction between a running neuronal network simulation and the environment, as well as interactive visualization, is being explored. This project, as well as the “Scalable brain-like cognitive architectures for virtual agents” project which is based at the KTH Department of Computational Science and Technology (CST) and run by Anders Lansner, Örjan Ekeberg, Pawel Herman and Christopher Peters, use MUSIC to connect simulations running on Milner with the outside world.