Closed ronentk closed 7 months ago
See https://wandb.ai/common-sense-makers/evaluation_benchmark/runs/gzxxek1w?
On the diagonal you see the true positives. Off diagonal You see the false positives as long as the row label predicted is a false negative.
This lets you see, when a label is not chosen correctly, which are the other falsely predicted labels that (might) have been predicted instead.
We do not use regular confusion matrix as this is a multi-label problem.
See https://wandb.ai/common-sense-makers/evaluation_benchmark/runs/gzxxek1w?
How to read the matrix:
On the diagonal you see the true positives. Off diagonal You see the false positives as long as the row label predicted is a false negative.
This lets you see, when a label is not chosen correctly, which are the other falsely predicted labels that (might) have been predicted instead.
Note
We do not use regular confusion matrix as this is a multi-label problem.