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

Neural network (NN) models are increasingly used in scientific simulations,AI, and other high performance computing (HPC) fields to extract knowledge fromdatasets. Each dataset requires tailored NN model architecture, but designingstructures by hand is a time-consuming and error-prone process. Neuralarchitecture search (NAS) automates the design of NN architectures. NASattempts to find well-performing NN models for specialized datsets, whereperformance is measured by key metrics that capture the NN capabilities (e.g.,accuracy of classification of samples in a dataset). Existing NAS methods areresource intensive, especially when searching for highly accurate models forlarger and larger datasets.  To address this problem, we propose a performance estimation strategy thatreduces the resources for training NNs and increases NAS throughput withoutjeopardizing accuracy. We implement our strategy via an engine called PEng4NNthat plugs into existing NAS methods; in doing so, PEng4NN predicts the finalaccuracy of NNs early in the training process, informs the NAS of NNperformance, and thus enables the NAS to terminate training NNs early. Weassess our engine on three diverse datasets (i.e., CIFAR-100, Fashion MNIST,and SVHN). By reducing the training epochs needed, our engine achievessubstantial throughput gain; on average, our engine saves 61% to 82% oftraining epochs, increasing throughput by a factor of 2.5 to 5 compared to astate-of-the-art NAS method. We achieve this gain without compromisingaccuracy, as we demonstrate with two key outcomes. First, across all our tests,between 74% and 97% of the ground truth best models lie in our set of predictedbest models. Second, the accuracy distributions of the ground truth best modelsand our predicted best models are comparable, with the mean accuracy valuesdiffering by at most .7 percentage points across all tests.

PEng4NN: An Accurate Performance Estimation Engine for Efficient Automated Neural Network Architecture Search