3 ms·
I wouldn't use "bias" here. It's a very ambiguous term in machine learning.
by cubefox 3y ago
I wouldn't use "bias" here. It's a very ambiguous term in machine learning.
- dafoex 3y agoNot sure what a better term would be... Just being more specific might help, like "bias in the training data"?
- cubefox 3y agoBut what does that mean? That the training distribution isn't representative of the population distribution, i.e. that it wasn't sampled IID? But the professors look pretty representative of real professors to me, so this doesn't sound very likely.