sega_sai

24 ms·

451.
by sega_sai 3y ago
For continuous data it is exactly a probability density evaluated on your data (for discrete it's PMF instead). L(params)=P(D|params)
452.
by sega_sai 3y ago
It is very strange that this is on a main page. The key thing is likelihood is probability density of your data! I.e. if your probability density is a Gaussian N(0,0.00001), then the likelihoods of data-points next to the mean will be ver
453.
by sega_sai 3y ago
That is creepy...
454.
by sega_sai 3y ago
Something's wrong with this. It seems it's a garbage in, garbage out rating. I.e. Edinburgh is at the top in the UK. I clicked there and it has a few measurements at 5 random places.
455.
by sega_sai 3y ago
There is this article https://www.intraway.com/blog/provisioning-volte But it seems that basically VoLTE is not enabled by default and each side needs to white list the other. The provider needs to white list the phone
456.
by sega_sai 3y ago
Yes, I think it is Google's responsibility as well. But it seems one of those situations, where user support is not Google's priority. I know that for Pixel 5 and later there is patch that you can apply (but it only works till the
457.
Google and a UK mobile operator cannot fix phone calls on Pixel devices(issuetracker.google.com)
3 points by sega_sai 3y ago|6 comments
458.
by sega_sai 3y ago
With 2G/3G being disabled in the UK, phones need to use VoLTE or VoWIFi to call. While Pixel phones from 4 to 7 support VoLTE it needs to be provisioned by the provider/Google. One of the UK providers Lebara cannot do that, pointi
459.
by sega_sai 3y ago
I think another way to think about it is to allow 'long planning' queries. I.e. where it is allowed to spend a second, or maybe a few seconds choosing the best plan. That may involve collecting more statistics or running a query f
460.
by sega_sai 3y ago
One nice illustration of "space is really big" is a fact that if you take the cube with the side equal to the distance from the Sun to the nearest star and fill it with water, then the mass of this cube will be roughly equal to th
461.
by sega_sai 3y ago
It is really cool. It would be interesting to make the speed proportional to the either the number of squares owned or the square root of that. Because right now as the area the ball owns gets smaller, the number of hits will increase at th
462.
by sega_sai 3y ago
Great method! It's better though to choose prime numbers for mod X operation, that'll make biases less likely.
463.
by sega_sai 3y ago
Because of the issues with OpenVPN/Wireguard blocking, a few months ago I completely switched to shadowsocks which I think mostly works. But it looks like https://github.com/amnezia-vpn/amneziawg-go -- is the way
464.
by sega_sai 3y ago
That's the same 'Capita' that had a breach last year: https://www.theguardian.com/business/2023/aug/04/cyber-attac... -- Why am I not surprised...
465.
by sega_sai 3y ago
From Andrew Gelman: "Any sufficiently crappy research is indistinguishable from fraud" https://statmodeling.stat.columbia.edu/2016/06/20/clarkes-la...
466.
by sega_sai 3y ago
Just in case the DOI link from the press-release doesn't work for you -- https://www.nature.com/articles/s41586-023-06747-5 here's the paper.
467.
by sega_sai 3y ago
I did not. I was applying from academia for quant position ( I got an offer but did not take it). Solving a bunch of K problems was one of homeworks.
468.
by sega_sai 3y ago
DB (10 years ago)
469.
by sega_sai 3y ago
I don't know, my personal/biased/limited impression from solving K language problems for the job interview is that the language is on purpose obtuse. It's a nice language for puzzles and clever solutions. But IMO, workin
470.
by sega_sai 3y ago
Maybe that's an unconventional take. But in my mind, the situation is the result of 'instantaneous' incentives in companies. I.e. CEOs and managers are rewarded when the stock price rises and are never penalised if things fal
471.
by sega_sai 3y ago
I wonder what happens when an unregulated market that can be easily manipulated is connected to a 'normal' stock exchange through instruments like Bitcoin ETF.
472.
by sega_sai 3y ago
The statement about 1965 is not correct if you take into account dividends. The longest periods of being down (in real terms) is 1973-1985, 2000-2013, 1937-1945, 1916-1925 ( https://observablehq.com/@tcgarvin/shiller-sp
473.
by sega_sai 3y ago
That is explicitly discussed in the text of the complaint. Google produces snippets with a direct link to the NYT and drives readers there.
474.
by sega_sai 3y ago
Any forecasts that don't involve probabilities or confidence intervals are useless. Also, if any forecasters were really serious, they would register their past forecasts and show how good they have been in the past. But I think most o
475.
by sega_sai 3y ago
Interesting. I wonder what are the two most separated pages, i.e. pages with the longest path connecting them.
476.
What's Going on with Google Assistant?(computerworld.com)
1 points by sega_sai 3y ago|0 comments
477.
by sega_sai 3y ago
Thank you for the link, it is an interesting read, and has a lot of funny quotes, stories and some interesting ideas. In the beginning I felt it was insightful, but by the end of the read I felt it was just a bunch of opinions based on pers
478.
Child's Online Mistake Can Ruin Your Digital Life(nytimes.com)
13 points by sega_sai 3y ago|3 comments
479.
by sega_sai 3y ago
Even if all the top 7s valuation goes to zero (which obviously will not happen) then the index will go down only by 29% which is not much for an equity based index, so I don't think it's that big of deal to be honest.
480.
by sega_sai 3y ago
This guy completely lost it. Nobody in the field treats him seriously because he publishes like 1 first paper a month with all sorts of speculations and shoddy science. Sure when his first claims were distributed about Oumuamua, people wer

More ›