Echo State Networks, part of the Reservoir Computing paradigm, are an efficient type of Recurrent Neural Network where the recurrent neurons only have a partial connectivity between themselves. Such networks where only a subset of all possible connections is made are known as sparsely connected networks. In Echo State Networks, only the weights from the hidden layer to the output layer may be altered during training.

Echo State Networks are useful for matching and reproducing specific input patterns. Since the only tap weights modified during training are the output layer tap weights, training is typically quick and computationally efficient in comparison to other multi-layer networks that are not sparsely connected.