Learning task-dependent distributed representations by backpropagation through structure

While neural networks are very successfully applied to the processing of fixed-length vectors and variable-length sequences, the current state of the art does not allow the efficient processing of structured objects of arbitrary shape (like logical terms, trees or graphs). We present a connectionist architecture together with a novel supervised learning scheme which is capable of solving inductive inference tasks on complex symbolic structures of arbitrary size. The most general structures that can be handled are labeled directed acyclic graphs. The major difference of our approach compared to others is that the structure-representations are exclusively tuned for the intended inference task. Our method is applied to tasks consisting in the classification of logical terms. These range from the detection of a certain subterm to the satisfaction of a specific unification pattern. Compared to previously known approaches we obtained superior results in that domain.