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 gaining in popularity as they are used to generatestate-of-the-art results for a variety of computer vision and machine learningapplications. At the same time, these networks have grown in depth andcomplexity in order to solve harder problems. Given the limitations in powerbudgets dedicated to these networks, the importance of low-power, low-memorysolutions has been stressed in recent years. While a large number of dedicatedhardware using different precisions has recently been proposed, there exists nocomprehensive study of different bit precisions and arithmetic in both inputsand network parameters. In this work, we address this issue and perform a studyof different bit-precisions in neural networks (from floating-point tofixed-point, powers of two, and binary). In our evaluation, we consider andanalyze the effect of precision scaling on both network accuracy and hardwaremetrics including memory footprint, power and energy consumption, and designarea. We also investigate training-time methodologies to compensate for thereduction in accuracy due to limited bit precision and demonstrate that in mostcases, precision scaling can deliver significant benefits in design metrics atthe cost of very modest decreases in network accuracy. In addition, we proposethat a small portion of the benefits achieved when using lower precisions canbe forfeited to increase the network size and therefore the accuracy. Weevaluate our experiments, using three well-recognized networks and datasets toshow its generality. We investigate the trade-offs and highlight the benefitsof using lower precisions in terms of energy and memory footprint.

Understanding the Impact of Precision Quantization on the Accuracy and Energy of Neural Networks