Key to "Saved Weights" file:

For each sensory node (ordered like characters on a page, first horizontally across the grid and then down to succeeding lines), the weights to the hidden layer are written on several separate lines, one for each bit in the "stimulus encoding" (these lines taken top-to-bottom represent sensors registering digits in the "stimulus encoding", read left-to-right). The default stimulus encoding, mentioned in the manual, has the first line reflecting weights from a "rock-sensor", the second from a "food-sensor", the third from a "predator-sensor", and the fourth from a "clear-sensor".

Each line contains the weights for the link from that sensor to the nodes of the hidden layer (where these are listed from left to right). The individual weight values are separated by three tab-characters.

As an example, suppose we have an organism with three hidden nodes (F G H), and five sensors (including the one in the "body") located as follows:

A
BCD
E

Then, if the stimulus encoding has not been altered, the weight for the link from the "rock sensor" at location E to the second hidden node, G, is found on line 17 (lines 17-20 are for location E, the fifth sensory node), in the second position (because we are interested in the link to the second hidden node).

At the end of the sensory-node-weights, two blank lines appear.

Then each hidden-layer node has its weights to the output nodes listed on a separate line (each weight is followed by three tab characters). The first line consists of links from the leftmost hidden node, the next line consists of links from the second-to-leftmost hidden node, etc. Also, the weights in each line in left-to-right order correspond to the output nodes in left-to-right order.