Our Research

Multistable perception

Certain visual displays are not perceived in a stable way but, from time to time and seeemingly spontaneously, their appearance wavers and settles in a distinctly different form. This phenomenon is called multistable perception and occurs with a variety of visual displays (as well as with ambiguous stimuli in the auditory and tactile domains). By studying the details of this perceptual dynamic, we can learn much about the nature of perceptual representations. For example, it seems that bistable perception reflects competing processes of stochastic integration at two levels of representation. Helping to interpret these results are computational models developed by our collaborators in Rome and Barcelona.


Structure-from-motion display

Our most recent papers on multi-stable perception:


Visual attention

Our long standing  interest are the perceptual consequences and neural correlates of visual attention. Here, we are particularly interested in visual experience outside the current attention focus. Currently, we are studying the relation between attention and motion grouping and the neural responses to stimuli outside the attention focus (using fMRI).


Enhanced brain activity under dual-task conditions

 

Our most recent papers on visual attention:


Neuromorphic VLSI

We collaborate with partners in Rome and Zurich on a hardware implementation of spiking neurons and plastic synapses (neuromorphic VLSI). Our immediate aim is to port established models of plasticity, learning, or memory to this platform and to thereby extend the applicability of neuromorphic devices. More philosophically, we believe that ‘building brains’ is a necessary precondition for ‘understanding brains.


Synapse layout on chip


Synapse circuit with Hebb-like plasticity


Simulation of neural populations interconnected by plastic synapses

 


Learning of goal-directed behaviour

More recently, we have begun to experiment on the learning of goal-directed behaviour and to interret the results with models of reinforcement learning. We are particularly interested in the susceptibility of habitual learning to temporal context. We suspect that this susceptibility is important for developing more flexible, goal-directed behaviour control.


Fractal patterns used in associative learning