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 excellent performance of recent self-supervised learning methods onvarious downstream tasks has attracted great attention from academia andindustry. Some recent research efforts have been devoted to self-supervisedmusic representation learning. Nevertheless, most of them learn to representequally-sized music clips in the waveform or a spectrogram. Despite beingeffective in some tasks, learning music representations in such a mannerlargely neglect the inherent part-whole hierarchies of music. Due to thehierarchical nature of the auditory cortex [24], understanding the bottom-upstructure of music, i.e., how different parts constitute the whole at differentlevels, is essential for music understanding and representation learning. Thiswork pursues hierarchical music representation learning and introduces theMusic-PAW framework, which enables feature interactions of cropped music clipswith part-whole hierarchies. From a technical perspective, we propose atransformer-based part-whole interaction module to progressively reason thestructural relationships between part-whole music clips at adjacent levels.Besides, to create a multi-hierarchy representation space, we devise ahierarchical contrastive learning objective to align part-whole musicrepresentations in adjacent hierarchies. The merits of audio representationlearning from part-whole hierarchies have been validated on various downstreamtasks, including music classification (single-label and multi-label), coversong identification and acoustic scene classification.

Music-PAW: Learning Music Representations via Hierarchical Part-whole Interaction and Contrast