We present an educational cortical neural network simulator for learning about, building, and testing the cerebral cortical layer structure. The user can learn about the cortical tissue structure by reading descriptions of neural cell types (pyramidal cells, basket cells, etc.) and studying default parameters as he/she designs the cortical structures by populating cell types in each cortical layer and specifying connections between cell layers. The simulator offers kids-friendly (fun, easy, and intuitive) interface, as is shown in the video below. In the performance mode, the operation of the neural network is displayed in semi-realistic 3D graphics.
In the demonstration, the audience can define their own cortical tissue structures and run the networks on an OCR database. Networks with Hebbian type learning rules such as Self Organizing Feature Maps can most easily implemented. The audience can also draw digits online and run the network. The network structures and test performances are recorded and the best performers are listed on the screen.
Michiro Negishi (Neuroverb)
1999: PhD in Cognitive and Neural Systems, Boston University 1999-2002: Postdoc, Rutgers University 2002-2016: Research Scientist, Yale University 2016-: CEO, Neuroverb