Backpropagation through structure

{{short description|Technique for training recursive neural networks}}

{{refimprove|date=May 2015}}

Backpropagation through structure (BPTS) is a gradient-based technique for training recursive neural networks, proposed in a 1996 paper written by Christoph Goller and Andreas Küchler.{{cite book|first1=Christoph|last1=Goller|first2=Andreas|last2=Küchler|title=Proceedings of International Conference on Neural Networks (ICNN'96)|s2cid=6536466|chapter=Learning Task-Dependent Distributed Representations by Backpropagation Through Structure|citeseerx = 10.1.1.49.1968|year=1996|volume=1|pages=347–352|doi=10.1109/ICNN.1996.548916|isbn=0-7803-3210-5}}