Metraipolis › archive › dc_a7bcfe89e796

From Masking to Merging: Rethinking SpecAugment for Efficient Audio Spectrogram Transformer — arXiv:2609.13260

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.13260
License (as archived): CC0-1.0
Archived: 2026-10-16T00:12:17.843Z
SHA-256 of the archived text: f7801d2991492a6954f3994e494da3bce47fcece1235e0e997419ce940ecbfa7
Stable URL of this page: https://metraipolis.com/documento/dc_a7bcfe89e796

How to cite

From Masking to Merging: Rethinking SpecAugment for Efficient Audio Spectrogram Transformer — arXiv:2609.13260 — arXiv (en), https://arxiv.org/abs/2609.13260 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_a7bcfe89e796 (Metraipolis archive dc_a7bcfe89e796, sha256 f7801d2991492a6954f3994e494da3bce47fcece1235e0e997419ce940ecbfa7)

BibTeX

@misc{dca7bcfe89e796,
  title = {From Masking to Merging: Rethinking SpecAugment for Efficient Audio Spectrogram Transformer — arXiv:2609.13260},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.13260},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_a7bcfe89e796; sha256 f7801d2991492a6954f3994e494da3bce47fcece1235e0e997419ce940ecbfa7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.13260. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

This paper proposes SpecAugment-Patch Merging, a simple yet effective method to accelerate Audio Spectrogram Transformer (AST) training. We first apply SpecAugment to mask input spectrograms at the patch level, and after positional embeddings are added, the method selects r pairs of masked patches and merges them, reducing the number of tokens processed by the Transformer. Increasing the number of merged pairs r from 0 to 100 keeps mAP on AudioSet nearly unchanged (34.07 to 34.08) while throughput increases from 43.3 to 49.3 samples/sec, which is a relatively 13.9% improvement. Similar patterns appear on ESC-50 and Speech Commands V2, where throughput steadily improves with only minor accuracy changes, demonstrating that this merging approach provides faster training with minimal performance loss.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-15 in cs.LG).
Categorie arXiv: Sound (cs.SD) ; Machine Learning (cs.LG); Audio and Speech Processing (eess.AS).
Pagina del lavoro: https://arxiv.org/abs/2609.13260 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-15T23:54:14.823Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-15?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).