HN user
Cross-entropy is not the KL divergence. There is an additional term in cross-entropy which is the entropy of the data distribution (i.e., independent of the model). So, you're right in that minimizing one is equivalent to minimizing the other.
https://stats.stackexchange.com/questions/357963/what-is-the...
The first slide makes the disclaimer that one is not "able to identify if an event was due to climate change", only likelihoods of certain events occurring under various conditions. People might treat those two things the same due to our mental heuristics, but they are not.
Obamacare won't fix it because it's a gutted compromise. Universal healthcare (which is the only way to get around private insurers not paying out, and not tying healthcare spending to income) is a non starter in this country and it's entirely due to one political faction. It's accurate to blame them for this.
Those colleagues were very wealthy by Indian standards to begin with and have only increased their purchasing power by being abroad. India does in fact have free care at government facilities and they are certainly not "low overhead and high skill". I'd pick rural American healthcare over rural Indian healthcare 10 times out of 10.
I have a PhD and this is absolutely not the reason universities don't want to hire their own. In the modern day, there is no reason whatsoever that university affiliation limits the spread of ideas or collaborations. The reason is there is a funnel - only the top x% of students get the TT positions and students of prestigious institutions are just more likely to have the big name publications that forms the sole criterion for academic hiring. Even the prestigious institutions like to spread the mantra of "we don't hire our own" but this is just so that professors can avoid having the awkward conversations with their students along the lines of "you're good, you're just not better than <y> candidate".
Is this subscription-based hardware though? It's an open linux device and you can write any software you want to work with it. There's no requirement to pay for their cloud storage if you don't want to.
Why is this an advantage? The factors in assessing the risk are the creditworthiness of the customer and the amount. Also, I find it hard to believe that the creditworthiness of a customer actually fluctuates fast enough for Affirm to be able to do some arbitrage on it.
I went to one of these liberal arts college. They all (typically) have science and math departments. Engineering typically doesn't fall under the purview of liberal arts (though pure science and math certainly do). The problem tends to be availability of upper-level electives if you actually care about learning certain topics and don't want to just complete the requirements. Also, your exposure to research opportunities are typically external to your own institution (there are exceptions to this). I'd say the required courses in the first two+ years are a better experience than large research universities though.
Not really. A lot of these small liberal arts colleges date back to the 1800s. The more prestigious ones are often priced like private R1 institutions as well. I'm sure there are some two-year institutions that started offering four-year degrees, but it's not typically the case afaik.
How is this relevant to the pro sports bring value through children's leagues discussion though? It appears these kids leagues exist completely independent of the professional teams, and would continue to do so even if the pro team left.
That's what these economic studies look at - is it worth paying for the stadium in order to prop up those businesses? The answer appears to be overwhelmingly "no".
By that logic the government should subsidize/fund Netflix, Amazon and Apple because they deliver and produce content that entertains me.
just how important big teams like say, Chelsea or Barca are for a huge chunk of not just the people who live in their home cities but also across the world.
Here "important to" is equivalent to "liked/entertained by", as a form of entertainment. The question here is what forms of entertainment should tax payers pay for and your comparison with art museums, operas and orchestras is apt.
let's say barca wanted to leave Barcelona, do you think the city shouldn't do anything about it because it helps billionaires?
If barca were extorting the city of Barcelona by demanding they pay 10s or 100s of millions to upgrade Camp Nou, then the city should tell the team to go to hell. This exact scenario has played out in multiple cities in the US and with the IOC and city bids.
And they visit, have a couple of beers, and leave. This influx is accounting for in the analysis of "is it worth it?" and the answer is a resounding "no". Well, it's a "no" for most large, dense cities, Green Bay could be interesting because it's not in a dense city. Green Bay is also a particularly bad example for this conversation, being a non-profit community owned team and all.
It is played, almost exclusively, by said immigrants from the Indian subcontinent. So only in a handful of cities where there's a critical mass of them that are somewhat recently immigrated (and there is room for grounds).
why cities keep on bidding for the olympics, which is otherwise hard to rationalise.
There was one bid for 2028 and one for 2032. Compare with 8+ bids in years past. There are fewer and fewer new cities that have never hosted before putting in bids (ostensibly due to insane construction costs/demands by the IOC). 3 out of 5 cities bidding for 2024 withdrew their bids and the other two got 2024 and 2028. 2032 seems to have gotten only one bid. What's happening is that this "intangible benefits" theory is dying as more and more of the world realizes that there are none. On the other hand the olympics and the world cup does seem like a good venue for authoritarian regimes to create spectacles.
Where are the companies that don't pull this shit?
Netflix? Amazon?
The people in the sports club will roll their eyes at you because [they think] you are a tin-foil hat wearing lunacy who is paranoid about surveillance from the boogeyman. They will refuse to switch their mode of communication that works for them and everyone else for one person (you) who is just causing trouble for no reason.
The vast majority of people don't think Facebook is the abhorrent service it is, and an even larger majority don't realize WA is owned by Facebook.
At a company there's a reporting chain and an HR department to ensure that even in this situation, the candidate isn't left ghosted - offer to shop the candidate's resume around and switch teams, or at the very least inform them. There is no valid excuse for a company ghosting an accepted candidate.
They need special livescribe dot pattern paper - any ol' paper won't work.
I purchased the remarkable 1 where there was nothing else that had the same latency for pens/styluses. The apple stylus is down to 9ms now, which is way faster than even the remarkable 2 (and that is amazing). I would be annoyed with a paperlike protector on an ipad, if I were trying to use it as an ipad, but for use solely as a notetaking device, it might be great.
Fair enough - the page change UX is a little clunky (and anything requiring changing screens many times on e-ink will be clunky). I read papers a lot more than I do textbooks. For fiction, the successive page turns are fine.
This was my primary use case for this and I love it - I need to be taking notes in the margins for me to stay engaged with the material. My rare pain points have been: * sometimes the resolution is not enough and if you really want to zoom into a figure, the UX to do that is a little clunky * colors can be critical for some figures and the papers/books don't take care to pick colors that are easily distinguishable in greyscale.
I place a lot of value on paperfeel for writing though (I tried writing on pdfs with an ipad and...ugh). I've also really come to appreciate the distraction-free setting (I thought I could just be disciplined and not need a device specifically for this - I was wrong). I also use it for note taking in meetings where I want to be present and pay attention the whole time - game changer there as well.
The comment you are replying to is not suggesting that Apple is as egregious as Microsoft. Shittiness can be measured on absolute terms - one doesn't have to accept the lowest bar as the standard.
you think google/fb production code runs on laptops?
They don't need to predict the actual sale price of every given property perfectly to make money. They need to be able to say that with 99% probability, we will find at least one offer for this house >= 500,000. The errant buyer with loads of cash is free to come in and offer more. As long as they turn a sufficient profit at the 500,000 price point they should be set. In competitive markets, for people familiar with local dynamics, this is very much possible in a short-term time window. The question is doing this at scale in a fully automated way for a much longer time horizon. Zillow tried and failed.
Any probability distribution of any attribute can be aribtrarily cut off at the 90th or 85th percentiles to make this point. Are you claiming there's something special that relates integer multiples of standard deviation of the IQ distribution to emacs usage?
"Everything was going perfect. I loved my job and the people I worked with, I just left because I am prone to doing that when there are no issues whatsoever."
The reality of not being able to publish negative results, and the publish or perish system of academia paralyzes researchers.