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 (DNNs) are state-of-the-art algorithms for multipleapplications, spanning from image classification to speech recognition. Whileproviding excellent accuracy, they often have enormous compute and memoryrequirements. As a result of this, quantized neural networks (QNNs) areincreasingly being adopted and deployed especially on embedded devices, thanksto their high accuracy, but also since they have significantly lower computeand memory requirements compared to their floating point equivalents. QNNdeployment is also being evaluated for safety-critical applications, such asautomotive, avionics, medical or industrial. These systems require functionalsafety, guaranteeing failure-free behaviour even in the presence of hardwarefaults. In general fault tolerance can be achieved by adding redundancy to thesystem, which further exacerbates the overall computational demands and makesit difficult to meet the power and performance requirements. In order todecrease the hardware cost for achieving functional safety, it is vital toexplore domain-specific solutions which can exploit the inherent features ofDNNs. In this work we present a novel methodology called fault-aware training(FAT), which includes error modeling during neural network (NN) training, tomake QNNs resilient to specific fault models on the device. Our experimentsshow that by injecting faults in the convolutional layers during training,highly accurate convolutional neural networks (CNNs) can be trained whichexhibits much better error tolerance compared to the original. Furthermore, weshow that redundant systems which are built from QNNs trained with FAT achievehigher worse-case accuracy at lower hardware cost. This has been validated fornumerous classification tasks including CIFAR10, GTSRB, SVHN and ImageNet.

FAT: Training Neural Networks for Reliable Inference Under Hardware Faults