BRN Podcast, page-408

Hi, the converted data is in the form of spikes (events) as input to the Akida neural fabric. The spike patterns may cause neurons to fire depending on whether a pattern is familiar and repeated (this works in a similar way in biology). When neurons fire their associated synaptic weights are adjusted because they've encountered a familiar pattern. The weights of synapses associated with a neuron (one-to-many relationship) are adjusted in a way that makes it more likely that that neuron will fire again if it encounters the same spike pattern. Synaptic weights represent the learning being done and as such they are the 'data' that Akida produces as output from the original CNN input and they can be saved in neural libraries as part of Akida's post processing. Hope this helps. AIMO.