Multi-class or multi-label in the Open Tasks description

Hello.
In the Open Tasks description,

  • Google Speech Commands is described as a multilabel classification problem, but it would be a multi-class classification.
  • NSynth is described as “Scene based multiclass prediction”, maybe it is a multi-label problem because of the following “categorization of a single note into one of 88 pitches, and 12 chromas (pitch classes).”

Excuse me if these are correct, and I’m missing something…

Hi @daisukelab ,

Thanks for your comments on the open-task descriptions.

  • You are correct, Google Speech Commands is a multi-class classification. We will update this in the documentation.
  • NSynth is also a multi-class classification problem. However you are correct that the description is a bit misleading. The predictor will only classify a single note into one of 88 pitches. The chroma that will be used for evaluation follows from the pitch – but the predictor will not explicitly label the chroma in addition to the pitch. We will update this description to be a bit more clear.

Thanks again for your feedback! Please don’t hesitate to reach out if you have any other comments or questions.

Best,
Jordie

1 Like

@daisukelab thank you for the feedback. Clarifications have been added to the NSynth and Google Speech Commands task descriptions.

See also 2021-06-07 updates.

1 Like