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

Deep neural networks are typically initialized with random weights, withvariances chosen to facilitate signal propagation and stable gradients. It isalso believed that diversity of features is an important property of theseinitializations. We construct a deep convolutional network with identicalfeatures by initializing almost all the weights to $0$. The architecture alsoenables perfect signal propagation and stable gradients, and achieves highaccuracy on standard benchmarks. This indicates that random, diverseinitializations are \textit{not} necessary for training neural networks. Anessential element in training this network is a mechanism of symmetry breaking;we study this phenomenon and find that standard GPU operations, which arenon-deterministic, can serve as a sufficient source of symmetry breaking toenable training.

Beyond Signal Propagation: Is Feature Diversity Necessary in Deep Neural Network Initialization?