4 ms·

Except that AI models, especially large deep ones, do NOT overfit like the author thinks. They exhibit what is now called "deep double descent" -- the validatio

by cs702 4y ago

Except that AI models, especially large deep ones, do NOT overfit like the author thinks. They exhibit what is now called "deep double descent" -- the validation error declines, then increases, and then declines again: https://openai.com/blog/deep-double-descent/ https://openai.com/blog/deep-double-descent/ A question I've pondered for a while is whether complex systems in the real world also exhibit double descent. For example, transitioning an online application that currently serves thousands of users to one that can serve millions and then billions requires reorganizing all code, processes, and infrastructure, making software development harder at first, but easier down the road. Anyone who has gone through it will tell you that it's like going through "phase transitions" that require "getting over humps." Similarly, startups that want to transition from small-scale to mid-size and then to large-scale businesses must increase operational complexity, making everything harder at first, but easier down the road. Anyone who has been with a startup that has grown from tiny two-person shop to large corporation will tell you that it's like going through "phase transitions" that require "getting over humps." Finally, it may be that whole countries and economies that want to improve the lives of their citizens may have to go through an interim period of less efficiency, making everything harder at first, but easier down the road. It may be that human progress involves "phase transitions" that require "getting over humps."

jaschasd 4y ago
Blog post author here. A brief note that I do discuss the deep double descent phenomenon in the blog. See the section starting with "One of the best understood causes of extreme overfitting is that the expressivity of the model being trained too closely matches the complexity of the proxy task." I avoided using the actual term double descent, since I thought it would add unnecessary complexity. Lesson learned for next time -- I should have at least had an endnote using that terminology!
- cs702 4y ago
  Thank you. As you probably know, the big deal about double descent is that once sufficiently large AI models cross the so-called "interpolation threshold" in training, and get over the hump, they start generalizing better -- the opposite of overfitting. State-of-the-art performance in fact requires getting over the hump. As far as I can tell, you did not mention any of that explicitly anywhere in your post. Also, all your plots show only the classical overfitting curve, not the actual curve we now see all the time with larger AI models like Transformers.
  - jaschasd 4y ago
    It's true that I don't go into detail about double descent, though I do describe how increasing capacity often reduces overfitting. I believe the figure labeled "Figure 1" illustrates what your are suggesting (despite being labeled Figure 1, it is actually at the bottom of the blog post, so maybe easy to miss).
    - cs702 4y ago
      > It's true that I don't go into detail about double descent, though I do describe how increasing capacity often reduces overfitting. I agree. > I believe the figure labeled "Figure 1" illustrates what your are suggesting (despite being labeled Figure 1, it is actually at the bottom of the blog post, so maybe easy to miss). Easy to miss, yes. I'm not sure it illustrates the phenomenon, though. That plot shows extreme overfitting (i.e., interpolation) by the 10,000 parameter model. No one really understands what actually happens after interpolation. There's in fact some anecdotal evidence that after crossing the interpolation threshold, large AI models trained with SGD gradually begin to ignore outliers and find simpler models (!) that generalize better (!). Counterintuitive, I know. This is an active area of research, with no good explanations yet, AFAIK.
      - jaschasd 4y ago
        (the plot shows extreme overfitting with a 10 parameter model, and interpolation with a 10,000 parameter model)
        cs702 4y ago
        Interpolation == extreme overfitting. Double descent phenomenon is what happens after interpolation. -- RESPONDING TO YOUR LAST COMMENT (after reaching thread depth limit): Think of it this way: Why and how does the model's performance continue to improve on previously unseen samples after the model has fully overfit (interpolated between) all training samples? Interpolation is not the end-point in training, but a temporary threshold after which models learn to generalize better, improving on interpolation. How is it that these models improve on interpolation?

by cs702 4y ago

jaschasd 4y ago

Blog post author here. A brief note that I do discuss the deep double descent phenomenon in the blog. See the section starting with "One of the best understood causes of extreme overfitting is that the expressivity of the model being trained too closely matches the complexity of the proxy task." I avoided using the actual term double descent, since I thought it would add unnecessary complexity. Lesson learned for next time -- I should have at least had an endnote using that terminology!

cs702 4y ago
Thank you. As you probably know, the big deal about double descent is that once sufficiently large AI models cross the so-called "interpolation threshold" in training, and get over the hump, they start generalizing better -- the opposite of overfitting. State-of-the-art performance in fact requires getting over the hump. As far as I can tell, you did not mention any of that explicitly anywhere in your post. Also, all your plots show only the classical overfitting curve, not the actual curve we now see all the time with larger AI models like Transformers.
- jaschasd 4y ago
  It's true that I don't go into detail about double descent, though I do describe how increasing capacity often reduces overfitting. I believe the figure labeled "Figure 1" illustrates what your are suggesting (despite being labeled Figure 1, it is actually at the bottom of the blog post, so maybe easy to miss).
  - cs702 4y ago
    > It's true that I don't go into detail about double descent, though I do describe how increasing capacity often reduces overfitting. I agree. > I believe the figure labeled "Figure 1" illustrates what your are suggesting (despite being labeled Figure 1, it is actually at the bottom of the blog post, so maybe easy to miss). Easy to miss, yes. I'm not sure it illustrates the phenomenon, though. That plot shows extreme overfitting (i.e., interpolation) by the 10,000 parameter model. No one really understands what actually happens after interpolation. There's in fact some anecdotal evidence that after crossing the interpolation threshold, large AI models trained with SGD gradually begin to ignore outliers and find simpler models (!) that generalize better (!). Counterintuitive, I know. This is an active area of research, with no good explanations yet, AFAIK.
    - jaschasd 4y ago
      (the plot shows extreme overfitting with a 10 parameter model, and interpolation with a 10,000 parameter model)
      - cs702 4y ago
        Interpolation == extreme overfitting. Double descent phenomenon is what happens after interpolation. -- RESPONDING TO YOUR LAST COMMENT (after reaching thread depth limit): Think of it this way: Why and how does the model's performance continue to improve on previously unseen samples after the model has fully overfit (interpolated between) all training samples? Interpolation is not the end-point in training, but a temporary threshold after which models learn to generalize better, improving on interpolation. How is it that these models improve on interpolation?