A Study of OWA Operators Learned in Convolutional Neural Networks

Ordered Weighted Averaging (OWA) operators have been integrated in Convolutional Neural Networks (CNNs) for image classification through the OWA layer. This layer lets CNN integrate global information about early stages, where most architectures only allow exploitation of local information. As a side effect this integration, becomes practical method determination operator weights, which is usually difficult task that complicates integration these other fields. In paper, we explore weights learned inside layer, characterizing them their basic properties orness and dispersion. We also compare to some families operators, namely Binomial operator, Stancu exponential RIM finding examples are currently impossible generalize parameterizations.