Standard neural networks trained with backpropagation algorithm are fully connected. Hypercube neurons have limited number of inputs. This is the implementation of network that is not fully conected and trainable with backpropagation algorithm.