eduzhai > Applied Sciences > Engineering >

dMelodies A Music Dataset for Disentanglement Learning

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 9 pages

Abstract: Representation learning focused on disentangling the underlying factors ofvariation in given data has become an important area of research in machinelearning. However, most of the studies in this area have relied on datasetsfrom the computer vision domain and thus, have not been readily extended tomusic. In this paper, we present a new symbolic music dataset that will helpresearchers working on disentanglement problems demonstrate the efficacy oftheir algorithms on diverse domains. This will also provide a means forevaluating algorithms specifically designed for music. To this end, we create adataset comprising of 2-bar monophonic melodies where each melody is the resultof a unique combination of nine latent factors that span ordinal, categorical,and binary types. The dataset is large enough (approx. 1.3 million data points)to train and test deep networks for disentanglement learning. In addition, wepresent benchmarking experiments using popular unsupervised disentanglementalgorithms on this dataset and compare the results with those obtained on animage-based dataset.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×