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

In the traditional deep compression framework, iteratively performing networkpruning and quantization can reduce the model size and computation cost to meetthe deployment requirements. However, such a step-wise application of pruningand quantization may lead to suboptimal solutions and unnecessary timeconsumption. In this paper, we tackle this issue by integrating network pruningand quantization as a unified joint compression problem and then use AutoML toautomatically solve it. We find the pruning process can be regarded as thechannel-wise quantization with 0 bit. Thus, the separate two-step pruning andquantization can be simplified as the one-step quantization with mixedprecision. This unification not only simplifies the compression pipeline butalso avoids the compression divergence. To implement this idea, we propose theautomated model compression by jointly applied pruning and quantization (AJPQ).AJPQ is designed with a hierarchical architecture: the layer controllercontrols the layer sparsity, and the channel controller decides the bit-widthfor each kernel. Following the same importance criterion, the layer controllerand the channel controller collaboratively decide the compression strategy.With the help of reinforcement learning, our one-step compression isautomatically achieved. Compared with the state-of-the-art automatedcompression methods, our method obtains a better accuracy while reducing thestorage considerably. For fixed precision quantization, AJPQ can reduce morethan five times model size and two times computation with a slight performanceincrease for Skynet in remote sensing object detection. When mixed-precision isallowed, AJPQ can reduce five times model size with only 1.06% top-5 accuracydecline for MobileNet in the classification task.

Automated Model Compression by Jointly Applied Pruning and Quantization