Secondary Structure Prediction: Helix, Sheet and Coil, which method is accurate? Which is more efficient?

Feed-forward neural networks have been used to predict protein secondary structure for about 2 decades. Feed-forward nets are trained using a set of patterns known as the training set for which the desired outputs are known in advance - a process known in the neural network literature as supervised learning. These feedforward nets take known sequence-structure pairs as training examples and try to generalize the concept of secondary structure for a given sequence.