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
Compression (with respect to a fixed training dataset) is by definition opposite to generalization. If you find the optimal compression dict to compress English Wikipedia it will obviously not be optimal for Spanish Wikipedia.
Because by definition compression discards all information that isn't relevant to the training goal, and by definition the training goal isn't what you want to generalize to (otherwise there is no generalization happening).
The most compressed model that does X is only capable of doing X. If it could do more then you could compress further.
I think I might be mistaking your point about compression with the "representation" needed to do good compression. It's just that for me if my brain is making skill programs to handle new situations, the ability to represent efficiently information seems like an important feature
Yeah, with respect to that I got a bit confused (don't mind me I am really slow..). So, if the generalization difficulty is some measure of the gap between the minimal time it takes to learn a useful skill program and how long it takes for that program to do well on novelty
Isn't this definition to a specific variation of what generalizion could be? I am thinking here something like, if I can compress information ok I loose information but don't I gain in terms of reach? I think I might be confusing this with making something like "universality"
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
