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

While state-of-the-art development in CNN topology, such as VGGNet andResNet, have become increasingly accurate, these networks are computationallyexpensive involving billions of arithmetic operations and parameters. Toimprove the classification accuracy, state-of-the-art CNNs usually involvelarge and complex convolutional layers. However, for certain applications, e.g.Internet of Things (IoT), where such CNNs are to be implemented onresource-constrained platforms, the CNN architectures have to be small andefficient. To deal with this problem, reducing the resource consumption inconvolutional layers has become one of the most significant solutions. In thiswork, a multi-objective optimisation approach is proposed to trade-off betweenthe amount of computation and network accuracy by using Multi-ObjectiveEvolutionary Algorithms (MOEAs). The number of convolution kernels and the sizeof these kernels are proportional to computational resource consumption ofCNNs. Therefore, this paper considers optimising the computational resourceconsumption by reducing the size and number of kernels in convolutional layers.Additionally, the use of unconventional kernel shapes has been investigated andresults show these clearly outperform the commonly used square convolutionkernels. The main contributions of this paper are therefore a methodology tosignificantly reduce computational cost of CNNs, based on unconventional kernelshapes, and provide different trade-offs for specific use cases. Theexperimental results further demonstrate that the proposed method achieveslarge improvements in resource consumption with no significant reduction innetwork performance. Compared with the benchmark CNN, the best trade-offarchitecture shows a reduction in multiplications of up to 6X and with slightincrease in classification accuracy on CIFAR-10 dataset.

Multi-objective Evolutionary Approach for Efficient Kernel Size and Shape for CNN