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/

Zendy Plus

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

Zendy Tools

Zendy Open

Application-specific integrated circuit (ASIC) implementations for DeepNeural Networks (DNNs) have been adopted in many systems because of theirhigher classification speed. However, although they may be characterized bybetter accuracy, larger DNNs require significant energy and area, therebylimiting their wide adoption. The energy consumption of DNNs is driven by bothmemory accesses and computation. Binarized Neural Networks (BNNs), as atrade-off between accuracy and energy consumption, can achieve great energyreduction, and have good accuracy for large DNNs due to its regularizationeffect. However, BNNs show poor accuracy when a smaller DNN configuration isadopted. In this paper, we propose a new DNN model, LightNN, which replaces themultiplications to one shift or a constrained number of shifts and adds. For afixed DNN configuration, LightNNs have better accuracy at a slight energyincrease than BNNs, yet are more energy efficient with only slightly lessaccuracy than conventional DNNs. Therefore, LightNNs provide more options forhardware designers to make trade-offs between accuracy and energy. Moreover,for large DNN configurations, LightNNs have a regularization effect, makingthem better in accuracy than conventional DNNs. These conclusions are verifiedby experiment using the MNIST and CIFAR-10 datasets for different DNNconfigurations.

LightNN: Filling the Gap between Conventional Deep Neural Networks and Binarized Networks