Bad data can also be the issue for high bias right?

{removed by mentor}

Before baseline performance was introduced we always considered that when training error and cross validation error are similar then the model is trainied well. But now it could also mean high bias which I understand but it could also be bad data right?

Since we are comparing the model with completely independent performance index like baseline performance it could be bad data or bad baseline index as well instead of just high bias right?

If the baseline is wrong, then I prefer to say the judegment of high bias is not reliable, rather than saying that it is a cause for high bias, because the conclusion of “high bias” is first questionable. If the data is bad, then I think it can be a reason for observing a high bias given a correct baseline.

PS: Looks like the screenshot is a quiz question? I think the slide is relevant to your question but the quiz itself is not, but I will have to remove the screenshot because we can’t post a quiz question publicly according to the code of conduct.

Yep, mb on posting quiz, I posted the quiz because I felt it is somewhat relevant since we have to answer based on Baseline, Jtrain and Jcv values