In the paper published in “Frontiers Neural Circuits”, Bard professor Arseny Khakhalin shows that a realistic artificial neural network, modeled after tadpole brain, can detect impeding collisions. In this study the network was not specifically designed or tuned for any particular task, but rather it was made to incorporate as much information about the tuning of actual neurons in real biological tadpole tecta as possible. After this realistic model was created, the team studied its properties in ways that would be hard to do in a real tadpole, and found that the network is uniquely suited to solve one of the key problems animals are facing: it naturally detects looming stimuli, and can help spatial navigation and predator detection.
Citation: Jang, E. V., Ramirez-Vizcarrondo, C., Aizenman, C. D., & Khakhalin, A. S. (2016). Emergence of selectivity to looming stimuli in a spiking network model of the optic tectum. Frontiers in Neural Circuits, 10.