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

Deep neural networks have achieved remarkable success in various challengingtasks. However, the black-box nature of such networks is not acceptable tocritical applications, such as healthcare. In particular, the existence ofadversarial examples and their overgeneralization to irrelevant,out-of-distribution inputs with high confidence makes it difficult, if notimpossible, to explain decisions by such networks. In this paper, we analyzethe underlying mechanism of generalization of deep neural networks and proposean ($n$, $k$) consensus algorithm which is insensitive to adversarial examplesand can reliably reject out-of-distribution samples. Furthermore, the consensusalgorithm is able to improve classification accuracy by using multiple traineddeep neural networks. To handle the complexity of deep neural networks, wecluster linear approximations of individual models and identify highlycorrelated clusters among different models to capture feature importancerobustly, resulting in improved interpretability. Motivated by the importanceof building accurate and interpretable prediction models for healthcare, ourexperimental results on an ICU dataset show the effectiveness of our algorithmin enhancing both the prediction accuracy and the interpretability of deepneural network models on one-year patient mortality prediction. In particular,while the proposed method maintains similar interpretability as conventionalshallow models such as logistic regression, it improves the prediction accuracysignificantly.

Consensus-based Interpretable Deep Neural Networks with Application to Mortality Prediction