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

The recently proposed Multilinear Compressive Learning (MCL) frameworkcombines Multilinear Compressive Sensing and Machine Learning into anend-to-end system that takes into account the multidimensional structure of thesignals when designing the sensing and feature synthesis components. The keyidea behind MCL is the assumption of the existence of a tensor subspace whichcan capture the essential features from the signal for the downstream learningtask. Thus, the ability to find such a discriminative tensor subspace andoptimize the system to project the signals onto that data manifold plays animportant role in Multilinear Compressive Learning. In this paper, we propose anovel solution to address both of the aforementioned requirements, i.e., How tofind those tensor subspaces in which the signals of interest are highlyseparable? and How to optimize the sensing and feature synthesis components totransform the original signals to the data manifold found in the firstquestion? In our proposal, the discovery of a high-quality data manifold isconducted by training a nonlinear compressive learning system on the inferencetask. Its knowledge of the data manifold of interest is then progressivelytransferred to the MCL components via multi-stage supervised training with thesupervisory information encoding how the compressed measurements, thesynthesized features, and the predictions should be like. The proposedknowledge transfer algorithm also comes with a semi-supervised adaption thatenables compressive learning models to utilize unlabeled data effectively.Extensive experiments demonstrate that the proposed knowledge transfer methodcan effectively train MCL models to compressively sense and synthesize betterfeatures for the learning tasks with improved performances, especially when thecomplexity of the learning task increases.

Multilinear Compressive Learning with Prior Knowledge