Mimic is a simple perceptron. It's purpose is to learn to fire a single output neuron when the corresponding input neuron is fired. The network may be trained automatically or manually.
Neurons are circles in various shades of pink. The brighter the shade, the more active the neuron is. Neurons which are not firing at all are black. Connections are represented by rectangles running between neurons. Positive connections are blue; negative connections are red. The stronger a connection, the thicker the rectangle. The Mimic network has four input neurons on the left, two hidden layer neurons in the middle, and four output neurons on the right.
The patterns that the network is to learn are shown in the four small boxes in the upper right of the applet. On the left of each box is an input pattern, on the right the output pattern which the network is to learn. Clicking on one of the boxes with the mouse copies that boxes input pattern into the network and propogates it through to the output. If the output is wrong, the network needs training.
To train the network manually, all the connection strengths must be adjusted by hand. To increase the strength of a connection, left-click on it. To decrease the strength, right-click on it. (If you only have one mouse button, click to increase and shift-click to decrease.) NOTE: Increasing the strength of a negative connection makes it less negative so the connection will appear smaller. Decreasing the strength makes it more negative so the connetion will appear larger. If you decrease the strength of a positive connection far enough, it will become negative. Increasing the strength of a negative connection sufficiently will make it positive Sometimes clicking on a connection makes too small a change to see the difference, clicking several times in a row should make the change visible. Also, there is a mimimum and a maximum width to the connections. Beyond those boundaries, changes in strength can not be seen but may still occur.
Clicking the train button cycles through the four input patterns and trains the network to produce the correct output for each. It may take several cycles to adequately train the network. At the end of each cycle, the changes in the connections may be seen.
If things get hopelessly stuck, click on the randomize button. This scrambles the connection strengths and allows you to start over.