3 ms·
It performs worse than baсkprop.
by LudwigNagasena 4y ago
It performs worse than baсkprop.
- qorrect 4y agoIs it slower and less accurate ? Or just slower.
- LudwigNagasena 4y agoIt is far less accurate compared to SOTA models. The paper says it should train faster, but it doesn't provide any metrics; so it's hard to make any "pound for pound" comparisons.
- UncleEntity 4y agoSo the initial “hey, this might be a good idea” implementation performs slightly worse than something that has had literally billions of dollars thrown at it?
- ekleraki 4y agoIt is slower to the same backprop that we have used for decades now. No comparisons to AdamW were made. In fact, this algorithm uses backprop at its core, but propagating through 0 layers.
- LudwigNagasena 4y agoThe question was "If it's not new or novel, why aren't people using it?". For example, take a look at this paper: https://arxiv.org/abs/1905.11786 https://arxiv.org/abs/1905.11786. It was published 3 years ago, it also does parallel layer-wise optimization, it even talks about being inspired by biology, though the objective function is different from Hinton's. Why aren't people using it? Because it performs worse. It is that simple. Is it an interesting area to explore? Probably. There are millions of interesting areas to explore. It doesn't mean it is worth using, at least yet.
- oneoff786 4y agoVery significantly worse.