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

The clear transparency of Deep Neural Networks (DNNs) is hampered by complexinternal structures and nonlinear transformations along deep hierarchies. Inthis paper, we propose a new attribution method, Relative Sectional Propagation(RSP), for fully decomposing the output predictions with the characteristics ofclass-discriminative attributions and clear objectness. We carefully revisitsome shortcomings of backpropagation-based attribution methods, which aretrade-off relations in decomposing DNNs. We define hostile factor as an elementthat interferes with finding the attributions of the target and propagate it ina distinguishable way to overcome the non-suppressed nature of activatedneurons. As a result, it is possible to assign the bi-polar relevance scores ofthe target (positive) and hostile (negative) attributions while maintainingeach attribution aligned with the importance. We also present the purgingtechniques to prevent the decrement of the gap between the relevance scores ofthe target and hostile attributions during backward propagation by eliminatingthe conflicting units to channel attribution map. Therefore, our method makesit possible to decompose the predictions of DNNs with clearerclass-discriminativeness and detailed elucidations of activation neuronscompared to the conventional attribution methods. In a verified experimentalenvironment, we report the results of the assessments: (i) Pointing Game, (ii)mIoU, and (iii) Model Sensitivity with PASCAL VOC 2007, MS COCO 2014, andImageNet datasets. The results demonstrate that our method outperforms existingbackward decomposition methods, including distinctive and intuitivevisualizations.

Interpreting Deep Neural Networks with Relative Sectional Propagation by Analyzing Comparative Gradients and Hostile Activations