Classless Association using Neural Networks

Federico Raue, Sebastian Palacio, Andreas Dengel, Marcus Liwicki

Nov 04, 2016 (modified: Jan 13, 2017) ICLR 2017 conference submission readers: everyone
  • Abstract: The goal of this paper is to train a model based on the relation between two instances that represent the same unknown class. This scenario is inspired by the Symbol Grounding Problem and the association learning in infants. We propose a novel model called Classless Association. It has two parallel Multilayer Perceptrons (MLP) that uses one network as a target of the other network, and vice versa. In addition, the presented model is trained based on an EM-approach, in which the output vectors are matched against a statistical distribution. We generate four classless datasets based on MNIST, where the input is two different instances of the same digit. In addition, the digits have a uniform distribution. Furthermore, our classless association model is evaluated against two scenarios: totally supervised and totally unsupervised. In the first scenario, our model reaches a good performance in terms of accuracy and the classless constraint. In the second scenario, our model reaches better results against two clustering algorithms.
  • TL;DR: Learning based on the relation between two instances of the same unknown class
  • Conflicts:,