AUTOMATIC TEACH-IN IS LIMITED TO PLANAR SURFACES ONLY
Research has highlighted the limitations of machine learning when applied to data that does not incorporate planar geometry. In 2016, a completely new discipline was discovered that is likely the answer to the limitations of NDCs. Deep learning allows scientists to work within a much broader theoretical framework where models can be learned on any geometric surface.
Deep learning comes from the construction of gauge convolutional neural networks. It is the result of a study conducted by Taco Cohen, Max Welling, Berkay Kicanaoglu, and Maurice Weiler at the University of Amsterdam and Qualcomm AI Research. The CNN gauge can detect patterns on spheres, in 2D arrays of pixels and on irregularly curved objects.
Welling even made the point: “This framework is a fairly definitive answer to the problem of in-depth learning on curved surfaces. “The NCCs for gauges, for example, offer more results for interpreting learning patterns in climate data. Algorithms can also bring innovation to the vision of UAVs and next-generation cars. They can detect patterns even on asymmetrically curved surfaces of objects such as brains or hearts.