Current models with no spatiotemporally variable input are designed to simulate spontaneous activity. It'll be useful to add the option to inject external input into the simulation given a specific spatiotemporal pattern. The main issue will be that such external input array will use a lot of CPU and GPU memory if we intend to have a separate input value for any given node at any given integration step. Unless the input is somehow compressed (e.g. define specific input shapes + onset times, then create the input array on-line while running the simulation).
Current models with no spatiotemporally variable input are designed to simulate spontaneous activity. It'll be useful to add the option to inject external input into the simulation given a specific spatiotemporal pattern. The main issue will be that such external input array will use a lot of CPU and GPU memory if we intend to have a separate input value for any given node at any given integration step. Unless the input is somehow compressed (e.g. define specific input shapes + onset times, then create the input array on-line while running the simulation).