Computational Neurophysics group
The group constructs cortical network models respecting the number of neurons in nature. This is achieved in the bottom-up approach by combining knowledge of anatomy and physiology into equations not tailored to a particular desired function. The natural density of the model networks removes any uncertainty about distortions of activity due to model size. The local structure of the cortex is conserved in evolution from mouse to human and is the same for sensory processing and motor planning. This dual universality raises hope that there are fundamental principles to discover. Verification with experimental data requires the development of simulation methods and the maintenance of software implementing the technology over decades. The group contributes to the generic open source simulation code NEST and promotes a culture where organizations and researchers view scientific software as infrastructure. The models are made available to the community such that in combination with reliable simulation code they can serve as building blocks for further studies and as platforms for the integration of hypotheses on brain function. The group contributes to neuromorphic computing by providing software reference systems for validation. Furthermore, the design of neuromorphic systems is seen as an integral part of brain research exposing limitations of knowledge and inspiring new research.
correlations in cortical networks, simulation technology, neuromorphic computing
Prof. Dr. Markus Diesmann
Building 15.22 / Room 4014
Correlations in Cortical Circuits
- Correlation structure in cortical networks.
- Active decorrelation by recurrent network dynamics.
- Effect of realistic network connectivity.
- Relation between microscopic and macroscopic network dynamics.
- Role of cell-type diversity.
- Role of realistic network connectivity.
Integrating CSA into NEST
- Reduction of complexity of NEST's naive connection routines through the use of CSA.
- Basing the topology module of NEST on CSA in order to obtain better scaling.
- Creating C++ implementation of CSA that can be used by neural simulators to describe network connectivity.