| Paper: | PS-1B.22 |
| Session: | Poster Session 1B |
| Location: | Symphony/Overture |
| Session Time: | Thursday, September 6, 18:45 - 20:45 |
| Presentation Time: | Thursday, September 6, 18:45 - 20:45 |
| Presentation: |
Poster
|
| Publication: |
2018 Conference on Cognitive Computational Neuroscience, 5-8 September 2018, Philadelphia, Pennsylvania |
| Paper Title: |
A selective diffusion model of brain network activity |
| Manuscript: |
Click here to view manuscript |
| DOI: |
https://doi.org/10.32470/CCN.2018.1195-0 |
| Authors: |
Daniel Graham, Yan Hao, Hobart & William Smith Colleges, United States |
| Abstract: |
Connectomics has made progress in elucidating the structure and functional significance of anatomic brain networks. Yet researchers are only beginning to consider design principles that support efficient routing on such networks. In engineered networks, packet switching was developed to efficiently support fast, asynchronous, sparse activity. However, little work has considered routing architectures that would be best suited to operational demands such as the need for reliability and to the constraints of sparse activity, low energy budgets, and small or non-existent node buffers. In this context, we constructed a selective diffusion model inspired by packet-switched architectures. We focus in particular on a model germane to visual system function, though we simulate activity on the entire macaque connectome. We use an agent-based system to address how the brain could trade off among sparseness, message loss, buffer size, and speed. We find that when nodes have no buffer, overall speed and population sparseness are maximal, but there is high message loss (>50%). Small buffers lead to modest reduction in speed, but drastically reduced message loss. However, population sparseness is comparable to physiological values across buffer size, suggesting that selective diffusion could be an efficient solution to brain-wide communication. |
