Deep SpatioTemporal Inference Network (DeSTIN) is a scalable deep learning architecture that relies on a combination of unsupervised learning and Bayesian inference. The paper DeSTIN: A Scalable Deep Learning Architecture with Application to High-Dimensional Robust Pattern Recognition by Itamar Arel, Derek Rose, and Robert Coop describes this method.
Briefly put, DeSTIN uses online clustering algorithms to hierarchically create cetroids in a way that loosely mimics the way humans understand things. It can be loosely classified as a Deep Learning Network (DLN).
DeSTIN source code and links to more papers are on GitHub at https://github.com/opencog/destin
Various notes & papers related to ways of implementing OpenCog with DeSTIN (DestinOpenCog)