3 ms·
For an HN audience, the "How Bestspotting cheated" post on Kaggle may be a better place to start: https://www.kaggle.com/bminixhofer/how-bestpetting-cheated htt
by paulgb 7y ago
For an HN audience, the "How Bestspotting cheated" post on Kaggle may be a better place to start: https://www.kaggle.com/bminixhofer/how-bestpetting-cheated https://www.kaggle.com/bminixhofer/how-bestpetting-cheated
- 7777fps 7y agoIt's an odd way to cheat too, if they realised they had data from the validation set, couldn't they have over-trained a model with the validation set in the training data?
- RandomWorker 7y agoAgreed, this actually happened in the buildings ASREA comp. the data was found online.
- paulgb 7y agoYes, that would have been harder to detect if it is allowed. My understanding is that for a kernel competition (like this one), you can't use model weights that you've trained outside the kernel. Oddly, I can't find a rule explicitly prohibiting it.
- koheripbal 7y agoIs it me, or has the HN user-sphere changes significantly this year. I feel as though comments are less informative, more judgemental, and chalk full of emotions. ...as if a younger crowd of users has started using it. Have the older-wiser HNers started using another site?