This is evident when watching networks learn using my visualisation program. The backpropagation algorithm creates spontaneous types of patterns which are put into or scuplted into the network briefely or for some period, and then there is a random shift to a new concept with a new pattern for its own breifity. This process repeats until there is a clear learned functon, at which point the random search abrubtly ends. These series of processes generate animations that are explsions of movement in particular directions and angles, composed of various lines and shapes. They are common among the entire landscape of the network and working in apparent unison, therefore making their appearence and trendy behaviour obvious. I’m going to be posting more videos.