Output details

A modal learning adaptive function neural network applied to handwritten digit recognition

<24> This paper represents the first time the snap-drift adaptive function neural network (SADFUNN), devised by the authors building on extensive research into both snap-drift and adaptive function neural networks, was applied to a very challenging pattern recognition problem. The results show that SADFUNN performs well in comparison to other established methods that require similar computational resources. The method is general and applicable to any pattern recognition or classification problem. SADFUNN is an advance on multilayer perceptrons; adopting the modal learning approach to combining the advantages of supervised and unsupervised learning, and utilising adaptive functions that can solve linearly inseparable problems in a single layer.