Presentation Details |
|
| |
| Name: |
 |
Big Data in Neuroscience: Where Is The Information? |
|
| |
| Time: |
|
Wednesday, June 25, 2014
09:00 am - 09:30 am |
|
| |
| Room: |
|
Hall 5
CCL - Congress Center Leipzig |
|
| |
| Breaks: | 07:30 am - 10:00 am Welcome Coffee |
| |
| Speaker: |
|
Joachim M. Buhmann, ETH Zurich |
|
| |
| Abstract: |
|
Neuroscience provides data at very different scale: microscopic imagery (ssTEM) measures information on the structure of neurons, diffusion tensor imaging tracks fibers between brain areas and the hemispheres, fMRI data measures spatial activity patterns to estimate dynamic causal models. At all three length scales, machine learning methods extract complex structures from vast amount of data. Model selection plays an important role in controlling the complexity of the explanation. Furthermore, the amount of supervision information is very scarce since human annotators display a high error rate on these ambiguous data sources. We advocate a maximum entropy approach to interpret such a data source and their model complexity is regularized by an information theoretic model selection principle. |
|
| |
|
;){kind=link}