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

Deep neural networks are extremely successful in various applications,however they exhibit high computational demands and energy consumption. This isexacerbated by stuttering technology scaling, prompting the need for novelapproaches to handle increasingly complex neural architectures. At the sametime, alternative computing technologies such as analog computing, whichpromise groundbreaking improvements in energy efficiency, are inevitablyfraught with noise and inaccurate calculations. Such noisy computations aremore energy efficient, and, given a fixed power budget, also more timeefficient. However, like any kind of unsafe optimization, they requirecountermeasures to ensure functionally correct results.  This work considers noisy computations in an abstract form, and gears tounderstand the implications of such noise on the accuracy of neural networkclassifiers as an exemplary workload. We propose a methodology called WalkingNoise which injects layer-specific noise to measure the robustness and toprovide insights on the learning dynamics. In more detail, we investigate theimplications of additive, multiplicative and mixed noise for differentclassification tasks and model architectures. While noisy trainingsignificantly increases robustness for all noise types, we observe inparticular that it results in increased weight magnitudes and thus inherentlyimproves the signal-to-noise ratio for additive noise injection. Contrarily,training with multiplicative noise can lead to a form of self-binarization ofthe model parameters, leading to extreme robustness. We conclude with adiscussion of the use of this methodology in practice, among others, discussingits use for tailored multi-execution in noisy environments.

Walking Noise: On Layer-Specific Robustness of Neural Architectures against Noisy Computations and Associated Characteristic Learning Dynamics