R.P.: Learning of invariant object recognition in a hierarchical network. Neural Networks (2014

Abstract. In this paper we propose an object recognition system imple-menting three basic principles: forming of temporal groups of features, learning in a hierarchical structure and using feedback for predicting future input. It gives very good results on public available datasets. Pre-condition for successful learning is that training images are presented to the system in an appropriate order such that images of the same object under similar viewing conditions follow each other. The system has mod-erate memory demands and a very big fraction of computing resources (during recognition) is spent on nearest neighbor search in a codebook of visual features, which can be sped up using locality sensitive hashing methods [1].