Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Notes on metrics' symmetry #343

Open
snk4tr opened this issue Feb 14, 2023 · 2 comments
Open

Notes on metrics' symmetry #343

snk4tr opened this issue Feb 14, 2023 · 2 comments
Labels
documentation Improvements or additions to documentation

Comments

@snk4tr
Copy link
Contributor

snk4tr commented Feb 14, 2023

Is your feature request related to a problem? Please describe.
Currently, it is not clear which metrics are symmetric and which are not. It may cause potential low reproducibility of reported metrics scores. The process of debugging such problems may be tedious and time consuming.

Describe the solution you'd like
All metrics have to include a note in the docstring on their symmetry. Moreover, it might make sense to add a warning for asymmetric metrics that would trigger every time a metric is run. @denproc @zakajd please comment on that.

Additional context
For PieAPP:
metric(distorted, reference) - 0.836
metric(reference, distorted) - 0.601

@snk4tr snk4tr added the documentation Improvements or additions to documentation label Feb 14, 2023
@denproc
Copy link
Collaborator

denproc commented Feb 14, 2023

Hi @snk4tr,
In general, our documentation explaines the inputs of the metrics and classes as input tensor and target tensor. Therefore, a user could simpy follow the explanation or use key words to provide arguments to avoid wrong results.

In my mind, adding a note to assymetric measures is a great suggestion to make the property more visible. As we improve the metric description, it is related to #318.

On the other hand, I would avoid adding proposed warning to the metrics. In this case, we have no additional information to warn user about potential errors due to our code, dependencies or sudden change of interface with a new release. The proper use of the metric is up to user and their ability to read the documentation.

@zakajd
Copy link
Collaborator

zakajd commented Feb 15, 2023

Agree with Denis.
Clearly stating the order of input and target tensors in documentation is important. Warning could become too annoying, so don't see this necessary.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
documentation Improvements or additions to documentation
Projects
None yet
Development

No branches or pull requests

3 participants