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 have been applied in many applications exhibitingextraordinary abilities in the field of computer vision. However, complexnetwork architectures challenge efficient real-time deployment and requiresignificant computation resources and energy costs. These challenges can beovercome through optimizations such as network compression. Network compressioncan often be realized with little loss of accuracy. In some cases accuracy mayeven improve. This paper provides a survey on two types of network compression:pruning and quantization. Pruning can be categorized as static if it isperformed offline or dynamic if it is performed at run-time. We compare pruningtechniques and describe criteria used to remove redundant computations. Wediscuss trade-offs in element-wise, channel-wise, shape-wise, filter-wise,layer-wise and even network-wise pruning. Quantization reduces computations byreducing the precision of the datatype. Weights, biases, and activations may bequantized typically to 8-bit integers although lower bit width implementationsare also discussed including binary neural networks. Both pruning andquantization can be used independently or combined. We compare currenttechniques, analyze their strengths and weaknesses, present compressed networkaccuracy results on a number of frameworks, and provide practical guidance forcompressing networks.

Pruning and Quantization for Deep Neural Network Acceleration: A Survey