Project Description:
The Multi-Attribute Task Battery (MATB) is a tool designed by NASA to test cognitive performance under load, specifically, how the human brain handles multitasking. To effectively study this tool and working memory as a whole, our team created a software that allows the users to define neural networks on either a 2D or a 3D brain model. On this model, the user can define nodes, modules, edges, and sensors, and modify parameters for the nodes and the overarching network to fit their specifications. Our software then trains the recurrent neural network on the MATB task before simulating an actual run (or multiple runs) of the MATB, which also has its own customizable parameters. This is to see how well the model performs on the task, considering neurological limitations such as context switching and ease of task. The simulation also includes visualizations, such as a heatmap, an EEG trace, and a power spectral density graph based on the user-defined sensors. Once the simulation(s) terminate, the results are exported to a summary file which contains information on the model, the brain, the training process, simulation statistics, and raw data that can be graphed or analyzed further.
