Postdoctoral Project: Architectural design principles for intelligence: Modularity vs. integration (synthetic)
Part of research project: Architectural design principles for intelligence - Modularity vs Integration
Description of the research project
This project will study how architectural design principles affect performance of an intelligent system. We will focus on the tradeoff between modularity and integration and compare this between a biological system (the human brain) and a synthetic system (a computational model). A key question is how the modularity of a cognitive system can be optimized for the tasks they have to solve. The project comprises two postdoctoral research positions, which focus on graph-theoretical analyses of functional connectivity of the brain, and the development of control-theoretic methods for optimizing visual network models, respectively. The former subproject will primarily supervised by J.-D. Haynes, the latter by J. Raisch and H. Sprekeler.
Description of the postdoctoral project
The postdoctoral researcher will develop control-theoretic methods for optimizing the architecture of visual network models, and will be primarily supervised by J. Raisch and H. Sprekeler. The subproject will comprise i) the development of a suitable mathematical theory, ii) simulations and the optimization of neural networks performing visual tasks, and iii) the analysis of the resulting data for comparison with data obtained from human neuroimaging.
Project start date: October 1, 2019 (an earlier starting date may be possible)
Please visit the website https://www.personalabteilung.tu-berlin.de/menue/jobs/stellenausschreibungen/, which gives you access to the legally binding German version.
- excellent mathematical skills,
- excellent programming skills,
- excellent English skills both written and spoken and
- a keen interest in intelligence research within an interdisciplinary and highly collaborative research team.
The ideal candidate holds a PhD in machine learning or control theory and has a strong background in training recurrent artificial neural networks.