3 ms·
> That is not even close to the same thing. It's a common interpretation: https://arxiv.org/abs/1706.06859 https://arxiv.org/abs/1706.06859
by sooheon 7y ago
> That is not even close to the same thing.
It's a common interpretation: https://arxiv.org/abs/1706.06859 https://arxiv.org/abs/1706.06859
- irq11 7y agoThere may be a paper on it, but it’s not a common view. In particular, this paper neglected to do the obvious thing: ensemble networks trained with dropout. It improves performance over dropout alone.