ANC Workshop Talk: Ioan Stanculescu and Matthias Hennig, Chair: David Sterratt

Ioan Stanculescu

Towards a hierarchical model for condition monitoring

The health condition of premature born babies receiving intensive is fragile and surrounded by uncertainty. It cannot be observed directly, but patterns in the monitoring data can be associated with physiological or artifactual factors.  Previously, the Factorial Switching Linear Dynamical System has been applied for inferring the filtering distribution of these events.

In this talk, I will discuss how we are developing the model in order to infer high-level clinical factors, such as neonatal sepsis. I will also show some of our work on slow trend detection.  

Matthias Hennig

Homeostatic control of neuronal network activity

Activity in neuronal networks is remarkably stable, and their performance is sometimes not substantially affected following even substantial insults such as stroke injuries or pathological changes due to Alzheimer's disease. This stability is maintained by homeostatic processes, dedicated mechanism that compensate for chronic changes in activity levels. Homeostatic processes have been intensively investigated in single neurons, but little is known about their function in networks. Here I will present recent results from a study of mean field models of recurrent networks, which point towards general principles underlying homeostatic control of network activity.