Kirill Shchegelskiy, BCCN Berlin

The discovery of the neurons that are active when an animal is in a particular spatial position (place cells) has inspired the creation of the numerous models that try to replicate their behaviour. Most of them use neural networks approach, in attempt to be realistic and computationally efficient at the same time. It is important to mention that these networks can be of various sizes and complexities, as well as the modalities and dimensions of the input spaces. Training the network and constructing a cognitive map from it are also nontrivial questions that depend not only on the model architecture, but also on the environment exploration strategies. In the proposed work one common place cells model was used to test the effects of the omnidirectional input properties and the exploration on the resulting cognitive map and the agent navigating performance. The simulations were done using the Stage robot simulator in several two-dimensional environments with obstacles of various forms and sizes.

Additional Information

Master thesis defence in the MSc Computational Neuroscience

Organized by

Verena Hafner / Robert Martin