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

Deep neural networks (DNNs) can be made hardware-efficient by reducing thenumerical precision of the weights and activations of the network and byimproving the network's resilience to noise. However, this gain in efficiencyoften comes at the cost of significantly reduced accuracy. In this paper, wepresent a novel approach to quantizing convolutional neural network. Theresulting networks perform all computations in low-precision, without requiringhigher-precision BN and nonlinearities, while still being highly accurate. Toachieve this result, we employ a novel quantization technique that learns tooptimally quantize the weights and activations of the network during training.Additionally, to enhance training convergence we use a new training technique,called gradual quantization. We leverage the nonlinear and normalizing behaviorof our quantization function to effectively remove the higher-precisionnonlinearities and BN from the network. The resulting convolutional layers arefully quantized to low precision, from input to output, ideal for neuralnetwork accelerators on the edge. We demonstrate the potential of this approachon different datasets and networks, showing that ternary-weight CNNs withlow-precision in- and outputs perform virtually on par with theirfull-precision equivalents. Finally, we analyze the influence of noise on theweights, activations and convolution outputs (multiply-accumulate, MAC) andpropose a strategy to improve network performance under noisy conditions.

FQ-Conv: Fully Quantized Convolution for Efficient and Accurate Inference