A simple technique for improving multi-class classification with neural networks

Thomas Kopinski (1) , Alexander Gepperth (2, 3) , Uwe Handmann (1)

1 University of Applied Sciences [Bottrop]
2 RV - Robotique et Vision
3 Flowers - Flowing Epigenetic Robots and Systems

We present a novel method to perform multi-class pattern classification with neural networks and test it on a challenging 3D hand gesture recognition problem. Our method consists of a standard one-against-all (OAA) classification, followed by another network layer classifying the resulting class scores, possibly augmented by the original raw input vector. This allows the network to disambiguate hard-to-separate classes as the distribution of class scores carries considerable information as well, and is in fact often used for assessing the confidence of a decision. We show that by this approach we are able to significantly boost our results , overall as well as for particular difficult cases, on the hard 10-class gesture classification task.

Domaines: Apprentissage [cs.LG]

Thomas Kopinski, Alexander Gepperth, Uwe Handmann. A simple technique for improving multi-class classification with neural networks. European Symposium on artificial neural networks (ESANN), Jun 2015, Bruges, Belgium. ⟨hal-01251009⟩