5 ms·
Is there a simple explanation of what the GCI means? Preferably without mathematical wording "for such a set that", "for a given function X where", "under a fr
by Chris2048 10y ago
Is there a simple explanation of what the GCI means?
Preferably without mathematical wording "for such a set that", "for a given function X where", "under a frobnobian measure if" etcetcetc
Bayes has simpler, word descriptions, for example.
- openasocket 10y agowhat's "the GCI" here? I don't recognize that acronym, I couldn't find any term in the article it could be, and googling doesn't really turn anything up...
- davidgrenier 10y agoSecond wikipedia link for CGI refers to computer-generated imagery. That or I didn't understand your question.
- n4r9 10y agoI think they must have commented on the wrong article by mistake. The "Gaussian Correlation Inequality" was recently proved, and discussed in this thread: https://news.ycombinator.com/item?id=13977554 https://news.ycombinator.com/item?id=13977554
- Chris2048 10y agoAh, you are right...
- lkrubner 10y agoAnd hopefully no one uses the phrase "...therefore, it is trivial to say that..."
- MaysonL 10y agoIf you have two convex shapes (like a circle and a rectangle) which have the same center, if you think of the center as a target, and throw darts at the target. The dart will be more likely to land in the intersection of the two shapes than would be derive by multiplying the probability of landing in one and the probability of landing in the other. If you were talking about independent events, like rolling 1s on two dice, you'd get the probability of both things happening by multiplying their independent probabilitis: so 1/6 * 1/6 = 1/36 probability for snake eyes. Another example would be the correlation between weight and height. If you took the median weight and median height, and took say the 10% of the population around that height and the 10% of the population around that weight, the overlap would be more than the 1% that would be calculated by taking 10% of 10%. The GCI says that for any reasonably distributed variables, this inequality (the probability of falling around the central tendency of all variable) will be greater than the product of the probability of falling around the center for each variable taken independently.
- Chinjut 10y agoTo be clear, this is presuming a (multidimensional) bell curve distribution for dart throws, or weights and heights, or for whatever. Let us be clear that the GCI is specifically about bell curve (i.e., normal, Gaussian, etc.) distributions, not about anything else. Also, I believe the shapes must be not only convex, but symmetric under reflection through the center. But it suffices to consider only parallelepipeds.
- andrepd 10y agoYou want to understand mathematics without understanding mathematics?
- Chris2048 10y agoI want an intuition of a concept in mathematics that isn't the same as the formal definition. I gave a specific example (Baye's) - being able to juggle the symbols is entirely different from interpretation. If you want the difference between an intuition and formalization, consider the "Copenhagen interpretation" which is non-mathematical, yet is based on mathematically formalised concepts.