Related Experiment Videos
Multirate method for co-simulation of electrical-chemical systems in multiscale modeling
Ekaterina Brocke1,2,3, Mikael Djurfeldt4, Upinder S Bhalla5
1Science for Life Laboratory, School of Computer Science and Communication, KTH Royal Institute of Technology, Stockholm, Sweden. ekaterina.brocke@scilifelab.se.
Journal of Computational Neuroscience
|April 9, 2017
Summary
This study introduces a novel multirate algorithm for multiscale modeling in neuroscience, enhancing computational efficiency and accuracy in co-simulations of brain processes like learning and memory.
Area of Science:
- Computational Neuroscience
- Multiscale Modeling
- Co-simulation
Background:
- Multiscale modeling is crucial for neuroscience but lacks a solid theoretical foundation.
- Existing co-simulation methods face challenges with time integration for systems with vastly different time scales.
- Efficient and accurate numerical methods are needed for complex brain system simulations.
Purpose of the Study:
- To develop a novel multirate algorithm for multiscale modeling in neuroscience.
- To address stability, accuracy, and efficiency challenges in time integration for co-simulations.
- To minimize communication overhead between model components in distributed computing.
Main Methods:
- Developed a new multirate algorithm for handling components with different time scales.
- Implemented recursive error estimation for individual component discretization.
- Focused on minimizing inter-component communication for co-simulation suitability.
Main Results:
- The multirate algorithm effectively manages components across diverse time scales.
- Recursive error control maintains numerical accuracy within acceptable bounds.
- Preliminary results demonstrate the potential of the multirate approach for complex simulations.
Conclusions:
- The proposed multirate algorithm offers an efficient computational framework for multiscale modeling.
- This method can significantly advance the theoretical basis and application of co-simulation in neuroscience.
- The approach shows promise for broader applicability beyond the tested model.