Searching for the orthogonal states of a neural network

Two approaches to finding orthogonal states of a neural network are discussed. The first is a recursive approach; it builds N orthogonal vectors form N/2 orthogonal vectors. The second is a formula approach, in which orthogonal vectors can be obtained using a formula. It is observed that the formula approach is more efficient than the recursive approach. Properties of the neural network constructed from the orthogonal vectors are presented. An example is discussed