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

Automated machine learning (AutoML) methods improve upon existing models byoptimizing various aspects of their design. While present methods focus onhyperparameters and neural network topologies, other aspects of neural networkdesign can be optimized as well. To further the state of the art in AutoML,this dissertation introduces techniques for discovering more powerfulactivation functions and establishing more robust weight initialization forneural networks. These contributions improve performance, but also provide newperspectives on neural network optimization. First, the dissertationdemonstrates that discovering solutions specialized to specific architecturesand tasks gives better performance than reusing general approaches. Second, itshows that jointly optimizing different components of neural networks issynergistic, and results in better performance than optimizing individualcomponents alone. Third, it demonstrates that learned representations areeasier to optimize than hard-coded ones, creating further opportunities forAutoML. The dissertation thus makes concrete progress towards fully automaticmachine learning in the future.

Optimizing Neural Networks through Activation Function Discovery and Automatic Weight Initialization