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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Convolutional Neural Networks (CNNs) have proven to be a powerfulstate-of-the-art method for image classification tasks. One drawback however isthe high computational complexity and high memory consumption of CNNs whichmakes them unfeasible for execution on embedded platforms which are constrainedon physical resources needed to support CNNs. Quantization has often been usedto efficiently optimize CNNs for memory and computational complexity at thecost of a loss of prediction accuracy. We therefore propose a method tooptimally quantize the weights, biases and activations of each layer of apre-trained CNN while controlling the loss in inference accuracy to enablequantized inference. We quantize the 32-bit floating-point precision parametersto low bitwidth fixed-point representations thereby finding optimal bitwidthsand fractional offsets for parameters of each layer of a given CNN. We quantizeparameters of a CNN post-training without re-training it. Our method isdesigned to quantize parameters of a CNN taking into account how otherparameters are quantized because ignoring quantization errors due to otherquantized parameters leads to a low precision CNN with accuracy losses of up to50% which is far beyond what is acceptable. Our final method therefore gives alow precision CNN with accuracy losses of less than 1%. As compared to a methodused by commercial tools that quantize all parameters to 8-bits, our approachprovides quantized CNN with averages of 53% lower memory consumption and 77.5%lower cost of executing multiplications for the two CNNs trained on the fourdatasets that we tested our work on. We find that layer-wise quantization ofparameters significantly helps in this process.

Fixed-point Quantization of Convolutional Neural Networks for Quantized Inference on Embedded Platforms