1. Values for a single input or target are in a ROW of the corresponding matrix.
2. Supervised NNs learn from paired input/target examples. If you have no target data, the best you can do is cluster the input data into similarity clusters.
Hope this helps.
Thank you for formally accepting my answer