You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Fun point: it's important to remember that, in machine learning, the problem you're trying to solve is *always* misspecified. You can only *optimize* (fit to the data you have), but your actual goal is to *generalize* -- and generalization is fundamentally ill-defined.https://twitter.com/mariohsouto/status/992242166121431040 …
The foundational tension in machine learning is between optimization and generalization. If you could define generalization as an objective, you could solve for it. But you can't! That's the point.
But… you can! Shorter programs/fewer rules/simpler Kolmogorov complexity/… means covering more with less, part of that more is in the training, part isn't → generalization That may be the main reason why physicists are "obsessed" with unifying (forces, theories,…)
No you still can’t, because free lunch theorems suggest that you *can not* define a loss that always generalizes.
Interesting. How do you think of the {loss, accuracy} gap between training and test set? An approximate indicator of generalization?
An indicator that you can't optimize for. (If you do, you're training on the test set.)
Too good point to remember
As I see it, whichever ML algorithm you use, there is a certain implicit generalization model involved. ML at its core is the optimization of the assumed generalization objective.
It's worth mentioning that we promote generalization implicitly via the the optimization objective when we use regularization, Bayesian methods, or simply choose the right hypothesis class (CNNs on images). Although it is true that gen-tion has no explicit closed-form expression.
Interesting. Does this follow from “no free lunch”?
In layman’s terms: if we could train an AI to design the positronic brain (which is inherently beyond our intellectual reach), we could avoid the AI apocalypse
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
