Slowness as a principle of learning:Search for fundamental working principles of the brain

Researchers simulate navigational scenarios

It has been known for some time that, during movement in a familiar environment, special nerve cells in the brains of rats, the so-called place cells, become active. Each place cell is only active in a small part of the environment, its place field. Place fields are the basis for spacial representation in the brain and make navigation possible. Computational neuroscientists at the Ruhr-University Bochum have run computer simulations of real-life experiments, in order to find out which basic computational laws are applied by the brain. The journal “Frontiers in Computational Neuroscience” has now published their findings. They show that slowness seems to play a major role as a fundamental processing paradigm of brain activity.

Real-Life experiments serve as basis for simulations

Neuroscientists Fabian Schönfeld and Prof. Dr. Laurenz Wiskott use six real-life experimental studies as a basis for their computer simulations. In these studies, rats navigate through different environments in their search for food. A special procedure is then used to create “activity maps”, which depict the place fields developing in the rodent’s brain. Then scientists observe how these place fields change, when the environments are manipulated, for example by removing or moving prominent cues, or by changing its shape.

Special algorithm uncovers slowness

The scientists use their own open source software, which is especially designed to replicate real-life experimental environments. In detailed virtual worlds a virtual rat explores the environment, while the software records its visual impressions. The acquired data is then processed with the help of an algorithm especially developed to detect slowness: the Slow Feature Analyis (SFA). The slowness principle states, that the more meaningful data changes slower in time, than less meaningful information. Fabian Schönfeld describes the process using an example: “Consider a noisy TV reception. The flickering noise changes much faster than the actual image. It can therefore be discarded without loss of meaningful information.” After the data has been processed in this way, activity maps are created, just like in the real-life studies.

Computer simulation reproduces results of real-life studies

Comparison of activity maps from real-life experiments and their simulations shows: the results match. The slowness principle seems to play a major role in the processing of visual input. In the rodent studies, as in the virtual experiments, some remarkable features of place cells can be found. Prominent cues serve as visual anchors. If these cues are rotated, the place field rotates with them. If they are removed, then place fields in close proximity become less active. The results also indicate that place fields are not discarded, when the shape of the environment changes. The brain uses prominent features as a reference and is able to scale and adapt existing place fields. “It is quite extraordinary that so many different experiments with a variety of results could be reproduced so successfully with a single algorithm based on slowness,” says Schönfeld.

Slowness is investigated further in Collaborative Research Center

The work of Schönfeld and Wiskott is funded by a grant from the Deutsche Forschungsgemeinschaft (German Research Foundation, DFG) for Collaborative Research Centre 874, which investigates the cognitive representation of sensory processes. With the help of this funding, a model for the behavior of place cells was developed that contributes to understanding how slowness can be a mechanism for the processing of sensory input in the brains of mammals. To investigate this further the researchers will team up with other scientists from the Collaborative Research Center to, for example, perform simulations of navigation in the dark.