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

Exploiting sparsity enables hardware systems to run neural networks fasterand more energy-efficiently. However, most prior sparsity-centric optimizationtechniques only accelerate the forward pass of neural networks and usuallyrequire an even longer training process with iterative pruning and retraining.We observe that artificially inducing sparsity in the gradients of the gates inan LSTM cell has little impact on the training quality. Further, we can enforcestructured sparsity in the gate gradients to make the LSTM backward pass up to45% faster than the state-of-the-art dense approach and 168% faster than thestate-of-the-art sparsifying method on modern GPUs. Though the structuredsparsifying method can impact the accuracy of a model, this performance gap canbe eliminated by mixing our sparse training method and the standard densetraining method. Experimental results show that the mixed method can achievecomparable results in a shorter time span than using purely dense training.

Structurally Sparsified Backward Propagation for Faster Long Short-Term Memory Training