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

Standard deep learning models that employ the categorical cross-entropy lossare known to perform well at image classification tasks. However, many standardmodels thus obtained often exhibit issues like feature redundancy, lowinterpretability, and poor calibration. A body of recent work has emerged thathas tried addressing some of these challenges by proposing the use of newregularization functions in addition to the cross-entropy loss. In this paper,we present some surprising findings that emerge from exploring the role ofsimple orthogonality constraints as a means of imposing physics-motivatedconstraints common in imaging. We propose an Orthogonal Sphere (OS) regularizerthat emerges from physics-based latent-representations under simplifyingassumptions. Under further simplifying assumptions, the OS constraint can bewritten in closed-form as a simple orthonormality term and be used along withthe cross-entropy loss function. The findings indicate that orthonormality lossfunction results in a) rich and diverse feature representations, b) robustnessto feature sub-selection, c) better semantic localization in the classactivation maps, and d) reduction in model calibration error. We demonstratethe effectiveness of the proposed OS regularization by providing quantitativeand qualitative results on four benchmark datasets - CIFAR10, CIFAR100, SVHNand tiny ImageNet.

Role of Orthogonality Constraints in Improving Properties of Deep Networks for Image Classification