English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

An iterative method of learning has become a paradigm for training deepconvolutional neural networks (DCNN). However, utilizing a non-iterativelearning strategy can accelerate the training process of the DCNN andsurprisingly such approach has been rarely explored by the deep learning (DL)community. It motivates this paper to introduce a non-iterative learningstrategy that eliminates the backpropagation (BP) at the top dense or fullyconnected (FC) layers of DCNN, resulting in, lower training time and higherperformance. The proposed method exploits the Moore-Penrose Inverse to pullback the current residual error to each FC layer, generating well-generalizedfeatures. Then using the recomputed features, i.e., the new generalizedfeatures the weights of each FC layer is computed according to theMoore-Penrose Inverse. We evaluate the proposed approach on six widely acceptedobject recognition benchmark datasets: Scene-15, CIFAR-10, CIFAR-100, SUN-397,Places365, and ImageNet. The experimental results show that the proposed methodobtains significant improvements over 30 state-of-the-art methods.Interestingly, it also indicates that any DCNN with the proposed method canprovide better performance than the same network with its original trainingbased on BP.

Non-iterative recomputation of dense layers for performance improvement of DCNN