4 ms·
I disagree with you. Weak baselines are inappropriate, especially when claiming 'state-of-the-art' results, which is very common in ML literature these days. If
by prashp 7y ago
I disagree with you. Weak baselines are inappropriate, especially when claiming 'state-of-the-art' results, which is very common in ML literature these days. If weak baselines are used as proof of superiority for a particular method, there are infinite other methods/minor adjustments which can be shown to be superior using the same logic.
- SubiculumCode 7y agoState of the art is already a dubiously vague expression, but I think you miss my point. I would think usually the point is not that it is superior approach, but a viable one to explore.