Enhanced Deep Learning with Improved Feature Subspace Separation
Mustafa Parlaktuna, Ali Sekmen, Ahmet Buğra Koku, Ayad Abdul-Malek · 2018
This research proposes a new deep convolutional network architecture that improves the feature subspace separation. In training, the system considers M classes of input sets {Ci}i=1Mand M deep convolutional networks {DNi}i=1Mwhose filter and other parameters are randomly initialized. For each input class Ci, Convolutional Neural Network generates a set of features Fi. Then, a local subspace Siis matched for each set Fi. This is followed with a full training of the deep convolutional network DNibased on a decision criteria developed with computation of rejections of all features in {Fi}i=1Mto Si. Five different deep convolutional network topologies are used to show that the proposed technique works better for small network topologies and has comparable performance to more complex