Interesting goodbye message...

@SimplyBeautifulArt I elaborated about the perceived problem in another room If you are interested, it is in there. But be aware that is the Anime room, so use the transcript if you don't want to be bugged.

@user21820 I guess I'll never understand some people? :/

@Derpy yeah o/

@SimplyBeautifulArt Which people?

SE people?

@user21820 people in general lol

@SimplyBeautifulArt Well, I can understand most people though. Most irrational behaviour like we have seen since last year is due to sheer arrogance.

Especially due to power.

What I can't understand is how it got to this point.

@user21820 yeah I haven't really been paying attention to that stuff tbh, but I can imagine that's along the lines of how it went

=) Just been minding my own business and writing my own software for root-finding lol

Hahaha nice.

I haven't really followed the whole business, but every time they pin some Meta post to the right-side bar, I see it and go −_−.

xD

have you seen my recent question about the geometric mean?

Recently they even silenced (mod-deleted) a highly-upvoted (something like +70) Meta post that was not even offensive; it had a very measured tone and as far as I recall did not say anything immoral (and you know I have quite a high moral standard)...

oof

It was on this thread, and even our moderators cannot see it because they don't have high rep on Meta.

@SimplyBeautifulArt I just looked at it. I didn't think about this issue before, since I never did much with IEEE floating-point. I either use integers or arbitrary precision. =D

But you're right; if the search space is exponentially distributed it makes sense to really cut it in half.

@user21820 rules really are a mess =(

An analogy is a school that claims that they want to stop bullying in school to justify them tying every student to their chair.

Very good; zero bullying (between students) now.

Well I don't think they'd force, or at least I hope they won't, the mods of the communities to strictly follow such things

Who knows. I hope not, of course.

@user21820 I do. Only post matching "about 70 upvotes" went thru a war that escalated to deletion. Originally some pieces of the post were removed, still things related to the pronoun arguments - then it went down to fighting and then deletion. To be clear, the ones who had to resort to delete were not the ones that originally edited the post to remove some paragraphs. The events are a little more complex than just deleted straight away.

@Derpy If you have a copy of the original, I'd like to take another look. I saw it when it was locked but not yet deleted, and if the original was bad then I might agree with the removal after all.

@user21820 haha understandable, I've always been doing things with integers or doubles unless I'm asking wolframalpha for help.

@Derpy: I'm fine with just the text in pastebin or something (no need markup).

@user21820 the thing is I'm honestly not sure sometimes. If one point is close to 0, the other point close to 1, and the root at say 0.3, the use of the geometric mean seems somewhat strange

@SimplyBeautifulArt That's why I said cut the search space, not "geometric mean".

As I put it in the link at the end of my answer, can one really expect the distribution of the answer to align with the distribution of the search space in such a scenario?

It is computationally troublesome and pointless to exactly halve the search space, but it is easy to do the following, which is better in my opinion: Simply binary search on the exponent and then on the mantissa.

yeah

This has all the good properties and none of the bad cases. =)

I won't post the original, sorry, but the thing started from a phrase that could be taken as "Neopronoun aren't real pronouns in English" (but I am not sure if that was the actual intent or what). Despite that, originally only that line was removed. Then it started a fuss with the users adding another line to say they were censored, community user removing it, them adding back, namecalling in the edits comments etc etc etc.

well

On top of that the post had multiple comments removed in bulk fashion, often getting legit comments caught in the middle of thing that deserved removal.

In response, the user namecalled an employee/mod. And that probably triggered the final deletion.

@user21820 The issue is still the same. Just because there's significantly more points on one side does not mean the root is more likely to be there. =/

Note, not trying to defend anyone here, just saying that that question went thru a lot of chaos so try to not assume things are as simpler as some says.

I've found that in many cases trying to do such a thing costs say 50% more iterations which could be avoided otherwise

I have lost a normal comment too, but I don't care because it is evident that in that case someone just did mass deletion out of hurry.

Hm well that's a shame @Derpy

The tire fire from the original Cellio incident was never doused, every time that topic returns fighting does to.

@Derpy Well since the version I saw seemed fine, I will stick to that evaluation until I see evidence to the contrary. I didn't assume it was a simple issue, by the way, but unfortunately cannot re-evaluate it without access to the original. I noticed that many people read what they want to read, so I can't rely on others telling me whether that post was offensive or not, except those I trust to be objective and reliable.

@user21820 oh, yep, I didn't meant that you shouldn't. I just meant that it was not deleted because of that phrase.

Thanks anyway for your summary of what happened.

The actual flow of events was phrase removed -> fighting -> deletion

You are free to think that the fighting was actually well deserved, I was just pointing out it was not straight deleted and that the two actions were performed by different users

@Derpy Which users? I consider all SE employees as a single entity.

(And I didn't know about the fighting so I can't judge whether it was good or not.)

Paragraph removal was made by employee staff, the deletion was done by an elected moderator

I see.

@SimplyBeautifulArt Why would it? If the values are close the exponent binary search takes few steps, and if the values are far then the exponent binary search is nearly optimal.

Anyway, if I may add, I would be far more concerned about Shog post and what he says in there.

@user21820 To be clear I am talking about hybrid methods which use a combination of binary search and interpolation to find the root, so the sooner the bracket becomes tight enough the sooner interpolation can be used to rapidly find the root. The way floats are structured, there are more floats near 0 than elsewhere, so it makes a big difference if one starts with [10^-10, 1] or [10^-1, 1]. I think it is more likely the case that the root tends to occur evenly when split with the arithmetic mean

but not with the geometric mean in this kind of case, even if the search space is skewed that way.

@Derpy yeah...

The fact an ex-employee says that specific line, it is on a whole different level of bad.

@user21820 perhaps also relevant, but the error formulas when interpolating simply do not care about the relative error at all, it's usually based on a product of the absolute errors of the given points

@Derpy What specific line? The 'bible' quote?

@user21820 since it is still there....

> ...And then all hell broke loose. Unwilling to review over a year's worth of history, management opted to pick a name and make an example, perhaps thinking that a sufficiently "strong" response would suffice to compensate for months of apathy.

@Derpy I didn't quite understand that line, actually. I suppose it was a reference for Monica, but I of course have no way to evaluate whether it is true or false, because up till now we have no public access to the TL transcript.

All we outsiders know is, something blew up.

Probably it is. I am just fairly worried that an ex employee is saying that.

@SimplyBeautifulArt I am not sure what you mean by "tends to occur evenly". If you want to find a zero to a certain precision, it means your search space is structure like floating-point, so an optimal binary search would need to do something like what I said. Notice that for each exponent the search space is linearly distributed so arithmetic bisection is actually optimal.

@Derpy Ahh I see what you're getting at. That's.. interesting. So the fact that they haven't removed it suggests that it is true? Lol!!

@SimplyBeautifulArt: Sorry typo: *structured like

Indeed, you will need to search more if your interval endpoint is closer to zero, for the very reason you point out; there are more possible answers near 0.

@user21820 I am not saying how you should read it, and you should decide what to think about it on your own. I just say it seems odd and quite bad if taken in context, and it comes from someone that was quite trusted from the userbase and found himself on the go with little to no warning from what I heard (see the fact that the community issued a fund-riser to help him while finding a new job)

@user21820 it's optimal in that sense, but when you combine it with other methods it might take too long to reach sufficient absolute error for other methods to take over

@SimplyBeautifulArt Ah so that is what you meant.

For example if you switch to Newton-Raphson when it is within quadratic convergence.

That's an interesting question. I'll have to think about it before I can say much.

@user21820 Right now I'm thinking maybe I should follow up any GM which fails to significantly cut down on the absolute error by using a more heavily weighted AM of (3x+y)/4 (where |x| < |y|) to "simulate" two iterations of AM

I think this should only fail to be better than just using AM in that situation once?

@TheSimpliFire yo o/

o/ @SimplyBeautifulArt