An animation of an 8-frame recognition trace in a 4-level Sparsey model. The model has 16 V1 macs, 4 V2 macs, and 1 V3 mac at the top. V2 and V3 macs are composed of 9 competitive modules (CMs) each with 9 cells.  The V1 macs consist of 16 CMs each with 12 cells (individual cells are somewhat hard to distinguish at V1). Macs are activated (red border) if they have a sufficient amount of bottom-up (U) afferent activity. Black/red cells are correctly/incorrectly activated cells; light green are incorrectly non-activated cells. On each frame, we show subsets of the active afferent U (blue), horizontal (H, green), and top-down (D, magenta) weights for all of the activated macs. It gets busy....and these are just subsets. The 8-frame input snippet depicts a 2-segment arm moving in a 16x16-pixel input space, which was presented once (single-trial learning) as part of a 14-snippet training set. A main point to see in this trace is that the accuracy of trace (fraction of the cells that are black) increases toward the end of the snippet, especially at the higher levels.

Ideally this animation should show the rapid movement of the U, H, and D signals between and within levels and the progressive updating of codes from V1 to V3 on each frame processed. We're working towards that goal.