The black dots and the lines represent the neurons and how they are connected. More on that here if you're interested. The maps have been trained on each data set independently, so you can see that they approximate the layout of the datapoints.
|
In this data set, there are five groups on both the left and right
side, and a one-to-one correspondence between the groups. Here is a
chart of the data, where the colors show the correspondence between
maps that will be learned by Hebbian connections. I'm sorry the
colors are hard to see; on the left map, start at the top left side of
the pentagon, this corresponds to the bottom group on the right map.
As you move clockwise around the pentagon, move up one group on the
right plot. The black dots and the lines represent the neurons and how they are connected. More on that here if you're interested. The maps have been trained on each data set independently, so you can see that they approximate the layout of the datapoints. |
|
|
| As before, we train the Hebbian connections by presenting associated data points to the two maps and make note of which neurons are activated at the same time. In the following five 2x2 charts, the post-training response to data presentation is shown. Five points, one from each region of the maps, are presented and activation on the opposite map is graphed gradiently. |
