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

Outlier detection (OD) has received continuous research interests due to itswide applications. With the development of deep learning, increasingly deep ODalgorithms are proposed. Despite the availability of numerous deep OD models,existing research has reported that the performance of deep models is extremelysensitive to the configuration of hyperparameters (HPs). However, the selectionof HPs for deep OD models remains a notoriously difficult task due to the lackof any labels and long list of HPs. In our study. we shed light on an essentialfactor, training time, that can introduce significant variation in theperformance of deep model. Even the performance is stable across other HPs,training time itself can cause a serious HP sensitivity issue. Motivated bythis finding, we are dedicated to formulating a strategy to terminate modeltraining at the optimal iteration. Specifically, we propose a novel metriccalled loss entropy to internally evaluate the model performance duringtraining while an automated training stopping algorithm is devised. To ourknowledge, our approach is the first to enable reliable identification of theoptimal training iteration during training without requiring any labels. Ourexperiments on tabular, image datasets show that our approach can be applied todiverse deep models and datasets. It not only enhances the robustness of deepmodels to their HPs, but also improves the performance and reduces plenty oftraining time compared to naive training.

Unleashing the Potential of Unsupervised Deep Outlier Detection through Automated Training Stopping