Informatik-Kolloquium: Supercomputational Neuroscience: understanding neural computation through high performance simulation
Mittwoch, 17.01.2018, 15.00 Uhr
Ort: Seminarraum 9222, im E3
Referent: Prof. Dr. Abigail Morrison, IAS-6 / INM-6 / SimLab Neuroscience, Jülich Research Center
The brain is different from any currently available technology. Each nerve cell receives inputs from thousands of other neurons and sends output to the same number of units. Furthermore, only 50% of the connections are established locally within a cubic millimeter of brain tissue. The remaining synapses are established with nerve cells distributed across the whole brain. Understanding the computational mechanisms of these large networks is a crucial step in developing brain-like information processing technologies, and the simulation of such networks is a necessary tool for this research. Only highly scalable simulation code, which enables efficient distribution of workload across multiple compute nodes and minimizes serial data structure overhead, can guarantee reasonable run-times. In the first part of my talk I will review the technology of the neuronal network simulator NEST (http://www.nest-simulator.org) and describe our memory modelling approach, which enabled us to extend the scalability of the software to the realm of the world’s most powerful supercomputers available for scientific research. In addition to scalability, recently implemented interfaces to non-neuronal applications have considerably extended the range of scientific questions that can be addressed by the simulator. In the second part, I will provide examples of insights into neural computation in both healthy and neurodegenerated conditions we gained using this technology, and conclude with an outlook on ongoing research to exploit the neuromorphic hardware platform SpiNNaker.
Es laden ein: Die Dozenten der Informatik