Training neural audio classifiers with few data

Descripció

• Dades relacionades

  https://github.com/jordipons/neural-classifiers-with-few-audio/

• Resum

  We investigate supervised learning strategies that improve the training of neural network audio classifiers on small annotated collections. In particular, we study whether (i) a naive regularization of the solution space, (ii) prototypical networks, (iii) transfer learning, or (iv) their combination, can foster deep learning models to better leverage a small amount of training examples. To this end, we evaluate (i–iv) for the tasks of acoustic event recognition and acoustic scene classification, considering from 1 to 100 labeled examples per class. Results indicate that transfer learning is a powerful strategy in such scenarios, but prototypical networks show promising results when one does not count with external or validation data.
  Llegir més + Llegir menys -

• Descripció

  Comunicació presentada a: ICASSP 2019 - 2019 IEEE International Conference on Acoustics, Speech and Signal Processing celebrat de 12 al 17 de maig de 2019 a Brighton, Regne Unit.

• DOI

  http://dx.doi.org/10.1109/ICASSP.2019.8682591

• Col·leccions

  Congressos (Departament de Tecnologies de la Informació i les Comunicacions)
Mostra el registre complet