3 ms·
I'm skeptical about both statements. I haven't heard the "comnon argument" you mention and I don't see how ignoring differences in distribution between laborato
by YeGoblynQueenne 9y ago
I'm skeptical about both statements. I haven't heard the "comnon argument" you mention and I don't see how ignoring differences in distribution between laboratory and real-world data will improve generalisation.
I find the last bit particularly hard to accept, but I'd like to hear why you say so, on both counts.
- chillee 9y agoSorry, my last sentence > Otherwise, models can achieve very good generalization. was probably unclear. When we talk about generalization, we typically have to assume a lack of distributional shift. Otherwise, if the "truly unseen data" really is drawn from a different distribution (say one where all cats are now called "dogs" and all dogs are now called "cats"), we can get arbitrarily poor performance. As for the flat minima part, lemme pull up some papers from my notes. https://arxiv.org/abs/1609.04836 https://arxiv.org/abs/1609.04836 showed larger batch sizes lead to sharper minima. > ... and as is well known, sharp minima lead to poorer generalization. https://arxiv.org/abs/1711.01530 https://arxiv.org/abs/1711.01530 provides a parametrization-invariant notion of "flatness" that counters Dinh et al's point in https://arxiv.org/abs/1703.04933 https://arxiv.org/abs/1703.04933. I thought this paper looks pretty promising, although I haven't gone through the math. >Although flatness of the loss function has been linked to generalization [Hochreiter and Schmidhuber, 1997], existing definitions of flatness are neither invariant to nodewise re-scalings of ReLU nets nor general coordinate transformations [Dinh et al., 2017] of the parameter space, which calls into question their utility for describing generalization https://arxiv.org/abs/1703.11008 https://arxiv.org/abs/1703.11008 is also a really promising paper on generalization that basically shows another way to view flatness from a Bayesian perspective (they also show the most impressive generalization bound I've seen yet). And of course, the classic Schmidhuber paper on Flat Minima: http://www.bioinf.jku.at/publications/older/3304.pdf http://www.bioinf.jku.at/publications/older/3304.pdf I think it's clear from these papers that the "flat minima generalize well" argument is a common avenue of research with respect to generalization. Many of the generalization papers I've read recently are mostly related to different ways of calculating flatness, and pretty much all of them make reference and link their work to the 1st and 4th paper I linked above.
- YeGoblynQueenne 9y agoThanks for the reply. What do you do, btw? I mean, where does your interest in this subject come from, if I'm not being nosy (again)? I'll have to have a look at those papers. Schmidhuber and Hochreiter first, obviously- my favourites. >> When we talk about generalization, we typically have to assume a lack of distributional shift. Otherwise, if the "truly unseen data" really is drawn from a different distribution (say one where all cats are now called "dogs" and all dogs are now called "cats"), we can get arbitrarily poor performance. Well, the difference in distribution doesn't have to be that radical, to lead to egregious mistakes. For example- I don't know if you caught wind of the recent "deep gaydar paper", where two researchers from Standford trained a deep learning classifier to identify gay men and women from images of their faces? The data in one experiment consisted of about 12k pictures of users of a US dating site. Shockingly (to me anyway), the number of pictures of gay men was exactly the same as the number of pictures of straight men, and so on for the numbers of pictures of gay and straight women [1]. This is a completely unrealistic, artificial distribution - estimates I've heard about the prevalence of homosexuality in the general population vary from 5% to 10%. The paper itself references a 7% prevalence. Given such artificial data, I really don't think anyone could train a classifier that would have the same ability to discriminate between categories when exposed to the real world. And we're not talking about "cats" looking like "dogs"- it's a simple binary problem, discriminating between gay/not gay, the same features are (presumably) associated with the same classes, except a classifier trained on balanced classes and in the real world the classes are unbalanced. Of course I can't really say what would happen if their classifier was trained on data with a distribution matching the real world (or their 7% expectation of gay men), but I do know that, with unbalanced classes, it's perfectly possible to train a classifier with perfect recall and rubbish precision (with a ratio of 10% gay men, you can get 1.0 recall with 0.37 ish precision). So with different training distributions you get different classifiers. I mean, duh. So that's the kind of thing I'm talking about. And for large enough problems (e.g. the language domain where the true problem is probably infinite) that is a real issue, because even with the best intentions and the best sampling techniques you may never get enough data to train a proper discriminator. And even if your laboratory experiments tell you that your discriminator is generalising well against the data you have, but didn't train it on, you have no way to tell what will happen when it encounters data that even you have yet to see. _________________ [1] That was for users with at least one picture, which is to say, everyone. Decreasing numbers of users had 2, 3 and so on up to 5 pictures and they were used in different experiments.