recursive neural network

Noun

 * 1)  A kind of deep neural network created by applying the same set of weights recursively over a structured input, to produce a structured prediction over variable-size input structures.