The Nineteenth Byte

LegionMammal978

00:25

Can anyone here recommend any good TeX editors? (It has to work with TeX Live installed on Ubuntu)

Maltysen

00:40

@LegionMammal978 i use vim :P

ASCII-only

02:08

@LegionMammal978 TeXstudio maybe

LegionMammal978

02:27

@ASCII-only Yeah, ended up reinstalling that, and then uninstalling the old TeX Live distros that apt installed with it

Also, does the size of a font measure the distance between the baseline and the cap height? Are these measured in DTP points, or some other point definition?

ASCII-only

02:41

> We are offer; - Word press Website/ Drupal website/ Magento Website/ Joomla website/ Ecommerce Website/Business Website/ Personal Website,

do these people offer normal websites :/

LegionMammal978

wat

Why does viewing a document at 100% scale not display it at 100% scale physically

ASCII-only

03:01

@LegionMammal978 nobody knows

New Main Posts

04:56

0

Q: SQL query shortest and longest name & alphabetical order (ORACLE)

Query the two cities in STATION with the shortest and longest CITY names, as well as their respective lengths (i.e.: number of characters in the name). If there is more than one smallest or largest city, choose the one that comes first when ordered alphabetically. Sample Input Let's say that CI...

sql

New Main Posts

06:11

-3

Q: what is error in this

include using namespace std; class Employee { private: int name; int age; public: void setName (int); void setAge (const int); int getName (); int getAge (); }; Employee::setAge (const int a) { age = a; } Employee::setName (int n) { name = n; } Employee:: getName() { retur...

random

Calvin's Hobbies

06:59

Would you ever pay $650 for a theatrical performance? Cause that's what Hamilton tickets are :(

Sherlock9

I might. Trouble is, since I'm on the wrong hemisphere, my list of expenses for watching Hamilton live starts with a round-trip plane ticket :(

Is the Chicago performance any cheaper?

feersum

09:05

lol, today on HNQ: "I want to make a fraction, but with no denominator or bar"

JUST THE NUMERATOR

Wen1now

09:58

Hi everyone

ASCII-only

hi

wait wat it's wen

what are you doing here

@feersum :/

Wen1now

Oh

ASCII-only

also. did that person refer to vinculum as bar.

prepares hunting gear

Wen1now

I was just wondering, does anyone have any recommendations on any golfing languages for me to learn

ASCII-only

~~Charcoal~~

Jelly/APL/J for people with a math background I guess

05AB1E/Actually would be suitable too I guess

flawr

10:08

@ASCII-only also MATL and Husk

ASCII-only

oh yeah i always forget about husk >_>

also kinda mthmtca

Wen1now

What, so all of them at once?

flawr

@Wen1now no, use Javagony first

user202729

@ASCII-only That language actually exist?

ASCII-only

@user202729 I did just link to it you know

user202729

10:12

@ASCII-only Only Jelly is actual golfing language in those.

ASCII-only

@user202729 of course. but the other two are still a lot golfier than most languages

user202729

@ASCII-only I see. Just that I didn't know about that and occasionally I see someone said that is is hypothetical.

Why does sometimes the star show as an empty star and sometimes a filled star?

dzaima

@user202729 empty stars are pinned by ROs

flawr

@user202729 a.k.a. stars without soul

user202729

So... there is [empty star] 0 or pinned message with 1 star is displayed as [empty star] 1?

ASCII-only

10:19

@user202729 pin = special star, so 1 star (just the pin and no other stars) displays as just a star...

@Mego fish is clearly superior to zsh (which is superior to bash)

><> >>>>>>>> fish but it's not a shell yet

user202729

Because I don't have those shells can anyone briefly tell me what benefits does those shells have over cmd?

ASCII-only

they're linux shells for one

user202729

So because linux is better than windows? Any other reasons?

ASCII-only

not that

but imo yeah the linux binutils can be used for much more than windows

also just cleaner syntax/way easier to use. especially with oh-my-zsh and oh-my-fish

feersum

10:37

I don't see why you guys are bashing cmd. What could be more convenient than the shortcut Alt-Space-E-P for pasting?

ASCII-only

@feersum where does that shortcut work

feersum

In cmd of course.

ASCII-only

@feersum if you mean cmd then ctrl+v and right click work too

it's not 1990 anymore

feersum

It is if you use Windows 7.

I'm assuming that must be a 10 "innovation".

ASCII-only

maybe

flawr

10:47

@feersum O_O never knew a shortcut existed, is there also a replacement for crtl+c? or crtl+x?

@feersum doesn't seem to work =/

feersum

@flawr What if you just press Alt-Space? Does anything happen?

flawr

Oh you don't have to press them simultaneously :)

(the drop downs are in german on my computer)

feersum

But you can.

flawr

doesn't seem to work here even with the german commands

ASCII-only

Alt has to be pressed before space though

user202729

10:54

@feersum Correct.

@ASCII-only No, Alt+Space, E, P.

ASCII-only

@user202729 huhwhat

Daruchini

hi guys

Mr. Xcoder

hello

user202729

(E stands for edit, P stands for paste) | For German commands it may be different.

Daruchini

Anyone here have experience with neural networks?

ASCII-only

10:56

@user202729 yeah? what about it?

@Daruchini ~~kinda~~ not really, why

Daruchini

I'm getting a bit confused trying to normalise continuous inputs

user202729

@Daruchini People often says "Just do it".

Daruchini

My model uses the sigmoid activation which has a domain of [0,1] so would that mean I should scale all inputs to fall within that range?

Cos I see loooads of repos who just use z-score norm

ASCII-only

Depends, I guess?

Daruchini

Really?

how so

user202729

10:59

@Daruchini I guess it would gives better results, but otherwise not scaling the input also works.

ASCII-only

I mean, z-score may be for inputs that don't really have limits?

@user202729 not scaling does not work very often

feersum

Isn't mapping stuff into the range [0, 1] is what sigmoids are for?

Daruchini

Yeh thats right @feersum , so would you ensure all inputs were mapped to that scale also

I'm just getting a little confused because I've seen multiple approaches

I know I could try them all and see which gives me better but wanted to know the math behind the approaches

ASCII-only

I'd say it depends on input distribution?

If you can't give hard limits, or almost all the inputs take up a very small part of the range then use z-score

user202729

@Daruchini I guess that it depends on which approach gives the "nicest" result function.

Daruchini

11:04

In my head, it makes more sense to try to keep the inputs within the active domain of the activation function being used. But I can see what @ASCII-only is saying

Plus i've looked at the density distributions of min-max normalisation and I can see you loose a bit of information

ASCII-only

You could also use z-score to just guarantee that the activation function gives reasonable results for most data you throw at it probably?

Daruchini

I don't understand?

ASCII-only

Like if you just scale to the z-score of a dataset you can basically guarantee the activation function works reasonably well no matter what dataset you pass it

Daruchini

Yeh I guess

I'll attempt both normalisation

user202729

11:51

@EricTressler While reading this PDF (background: because I find my algorithm for the next OEIS sequence too slow) I notice that it mentions level of a poset, even when it is not necessarily graded. What is it, or I just misunderstood the sequence and the poset are graded?

Simply Beautiful Art

12:14

Where can I get golfscript assistance?

New Main Posts

0

Q: Encode and Decode me

Given any input, you need to do is either decode or encode it in or from BASE16,32,64. So your task if you choose to accept will be to make a base16,32,64 encoder,decoder RULES: Standard Loopholes apply It should not be a builtin function. It must be a full program and that's all INPUT: In any ...

code-golf

Simply Beautiful Art

In particular, I'm trying to write a function with multiple variables.

Hm, I guess I want to take the second item on the stack.

But I'm not sure how to call it.

New Sandboxed Posts

12:44

0

A: Sandbox for Proposed Challenges

I asked this question in main site. Got a very bad review. So, I would be glad if any kind soul here would like to help me make this question better. And Some feedback on how can i make it good. Given any input, you need to encode / decode it in or from base 16, 32, 64. So your task will be to ...

feersum

Don't you hate it when you spend a while writing a test for something, and then it works the first time?

All that testing was for nothing!

ASCII-only

@feersum it's so it doesn't break later on

feersum

Not if it's a single-use piece of code :P

flawr

14:04

You can recognize a mathematician in that he always checks his Sudoku is correct because he knows that at this point he has only proved unicity :) — François Brunault Apr 29 '13 at 20:45

Mr. Xcoder

It's the first time I ever use an RNG for test cases :-D

New Meta Posts

14:20

0

Q: fixed header bar

I want a fixed header bar in PPCG (like in SE), but my custom css doesn´t work any more with the new layout. (also, I cannot remember where I put it - in Firefox. It was not in userContent.css) this does not work (excerpt from userContent.css): @-moz-document domain(stackexchange.com) { .to...

support css

orlp

15:06

CMC: make a Python expression that results in the integer 2 using only the characters [<]

feersum

@orlp [[]<[[]]][[]<[]]<<[[]<[[]]][[]<[]]

orlp

@feersum that's exactly what I had :)

feersum

There's not a ton of choice!

orlp

especially using [x][[]<[]] just to parenthesize x feels weird

but it works :P

@feersum I had an interesting problem I've been kinda stuck on for a while (although I didn't spend too much time on it)

given a list of positive elements we can 'split' an element x by removing it from the list and adding a, b to the list such that a + b = x

NieDzejkob

is the same possible for 3?

feersum

15:13

Only powers of 2, I would expect.

user202729

Because << is leftshift.

NieDzejkob

I know

orlp

@feersum what is an optimal algorithm that given a list l and n splits minimizes the maximum in the resulting list

feersum

Was this a recent question?

orlp

I've stated it in primes and squares a while ago

n = 1 is trivial, splitting the maximum element of l into two equal splits

but even n = 2 already things can start getting more difficult

user202729

15:16

I don't understand the question?

orlp

so say we have a list [1, 3]

muddyfish

@orlp [[]<[[]<[]]][[]<[]]<<[[]<[]][[]<[]]

1 byte more but gives 1 instead

orlp

we can pick an element and split it however we want

feersum

If we have a single number n, and k splits, we turn it into n / (k + 1), right?

user202729

If you split n times you'll get n+1 sublists, which maximum do you want to minimize?

orlp

15:17

in this case it's optimal to first split the 3 in 1 + 2

feersum

So it seems we can binary search on the answer

orlp

[1, 1, 2]

and then the 2 into another 1+ 1

[1, 1, 1, 1]

but this shows that splitting in half isn't always optimal

user202729

Ah you meant replace an element with a sum of two integers?

orlp

not integers

feersum

And see whether the number of splits required to get below x is greater than the amount we have.

orlp

15:17

numbers

NieDzejkob

@muddyfish [[]<[[]]][[]<[]]<<[[]<[]][[]<[]] is shorter

user202729

So you can split 5 into 2.5, 2.5?

orlp

@user202729 yes

user202729

@feersum Binary search seems to be correct.

orlp

@feersum I believe so

let's say the length of the list is n, and the number of splits is k

we've solved k <= 1

and I believe that what you say is correct, solving n = 1

Erik the Outgolfer

15:19

mmm calculator golf...nice

orlp

@feersum the difficulty is that you can have multiple 'barriers'

e.g. if the input is [1, 1, 1, 1, 5, 5, 5, 8, 12]

user202729

@orlp Just take the sum of the number of necessary split for each number?

feersum

What do you mean? Simply ask whether the amount of splits required to get the minimum below x is <= k.

orlp

now depending on k you can reach certain minima, but 1 and 5 are clearly 'barriers' that you need to break

user202729

@orlp I don't understand.

feersum

15:22

So you start with the interval [0, 12]. Check can you get down to 6? then continue with next interval.

orlp

@feersum that makes sense but I'm not convinced that it's always optimal

effectively what you're saying is that there's some minimal value v we can reach with k

feersum

Of course there is a minimal value.

orlp

and that we just handle each number individually to get to v

Mr. Xcoder

@LuisMendo MWAHAHA, tied MATL for this challenge by abusing a bug hehe :-)

feersum

Yes, there is no interaction between the groups fissioned from different starting numbers.

orlp

15:29

so if we have a certain conjectured minimal value v we need to check if sum(ceil(x/v - 1) for x in l) <= k

@feersum interestingly, since we're dealing with real numbers this process doesn't have to be finite

feersum

It is with one simple trick.

You update your binary search range with the number you actually found

Rather than the one you tested for.

orlp

what do you mean

the number you actually found?

oh I see

feersum

The actual min of x / ceil(x/v)

Mr. Xcoder

I have [[]<[[]]][[]<[]]<<[[]<[[]]][[]<[]], kinda late to the party. EDIT: This seems identical to feersum's.

orlp

@feersum don't you mean max?

feersum

15:36

@orlp Right.

Mr. Xcoder

@muddyfish You know you can produce 1 with []<[[]] don't you?

orlp

@feersum this is not enough

you can only do that if sum(ceil(x/v - 1) for x in l) <= k holds

feersum

Oh you mean if the binary search condiiton has the same value every time?

orlp

def minmax(l, k):
    lo = 0
    hi = max(l)
    while abs(hi - lo) > 0.01:
        mid = (lo + hi) / 2
        needed = sum(math.ceil(x / mid) - 1 for x in l)
        actual = min(x / ceil(x / mid) for x in l)
        print(lo, hi, actual)
        if needed <= k:
            hi = actual
        else:
            lo = mid
    return lo

this is what I have at the moment

and you can see what it prints

>>> minmax([1, 3], 2)
0 3 1.0
0 1.0 0.5
0.5 1.0 0.5
0.75 1.0 0.5
0.875 1.0 0.5
0.9375 1.0 0.5
0.96875 1.0 0.5
0.984375 1.0 0.5

the 'actual' maximum it found is 0.5

user202729

Probably the actual code is bugged.

feersum

15:44

@orlp Don't you mean max instead of min? :)

orlp

@feersum eh I had that just a second ago (was second guessing myself)

>>> minmax([1, 3], 2)
0 3 1.5
0 1.5 0.75
0.75 1.5 1.0
0.75 1.0 0.75
0.875 1.0 0.75
0.9375 1.0 0.75
0.96875 1.0 0.75
0.984375 1.0 0.75

feersum

No really, you made the exact same mistake as in my earlier message.

The line with actual = ...

orlp

yes

that's what I mean

I had max there

I was trying min because I was second guessing myself on the correction

the second output I posted is with max there

user202729

I guess you need to switch the if?

orlp

@user202729 no, the if order is correct

user202729

15:47

If the number of split needed for mid is < the allowed number of split should you look for a number > mid (so lo = mid)?

feersum

How are you getting 0.75? That should be impossible.

orlp

@feersum it is impossible, that's my point

when needed > k we get impossible results

feersum

Oh, it always prints actual even if the test failed.

orlp

actual is just the best result you'd get if you could actually use needed amount of splits

feersum

So this is indeed a problem, adversarial examples where the test always fails.

orlp

15:50

and that's why I have this

        if needed <= k:
            hi = actual
        else:
            lo = mid

if needed is permissible, sure we can use the actual value we calculated as you suggested

but if it's not we can get significantly better results (read: bogus)

feersum

Right, but it doesn't help if the else branch is taken every time.

orlp

@feersum actually

when needed == k we can just return actual

but I'm not convinced we'll always reach that spot

feersum

What if we just throw in a test of the upper endpoint

Alternatively, add a random number to the starting value of hi.

Then the probablity of getting stuck is 0.

New Main Posts

-1

Q: How google determines the OS and CPU bit of user machine?

When downloading Google Chrome, website is able to determine the user's operating system and whether PC is running on 32 bit or 64 bit. How does it do that?

functional-programming

2

Q: An Array of Challenges #3: Moving Averages

Note: This is #3 in a series of array-manipulation challenges. For the previous challenge, click here. Moving Average of a List The moving average of a list is a calculation resulting in a new, smoothed out list, created by averaging small overlapping sublists of the original. When creatin...

code-golf math arithmetic array-manipulation statistics

orlp

@feersum actually

the real issue here is the lo/hi assumption

by setting lo = 0 we've already allowed invalid solutions in our [lo, hi] range

hi will always be valid

feersum

15:58

Doesn't matter, you will never actually test one of the endpoints.

Mr. Xcoder

Could you please vote to close this?

orlp

@feersum but suppose in the case where hi is the optimal solution

how do we detect this scenario?

feersum

I think either of my two solutions above could work.

orlp

it appears that checking needed == k works

actually that doesn't work

case in point: minmax([1, 3], 3)

because minmax([1, 3], 2) == minmax([1, 3], 3)

user202729

@Mr.Xcoder Is the comment below auto-generated?

feersum

16:01

Checking needed == k would be correct, but hardly necessary.

orlp

@feersum what do you mean by this

feersum

At each step, check hi before checking mid. You'd also need the needed == k for this actually.

user202729

I guess it's not hard to check if hi is the optimal solution.

caird coinheringaahing

@user202729 Yeah, if you post a custom close message, it auto-converts it to a comment, and people who close with that reason auto-upvote the comment.

feersum

I like the random number solution more though.

Mr. Xcoder

16:04

@user202729 Yes.

orlp

@feersum what exactly would you want to check hi with?

user202729

I guess the condition is, for all x in the list, $\lfloor \frac x {lo} \rfloor = \frac x {hi} $.

feersum

@orlp Same thing you check with mid.

caird coinheringaahing

@NewMainPosts ಠ_ಠ I spent 5 minutes working on that before remembering the sublists builtin in Jelly

orlp

actually

nevermind...

@feersum but we already know that hi is always valid, that is the needed check for hi always passes

feersum

16:07

@orlp That's why you need the needed == k exit condition.

Eric Tressler

@user202729 I'm not really sure, unless it was being used informally, or if level means something else in context. And I don't see the example you're referring to

user202729

@EricTressler Ok no problem I've got help on mathematics chat room.

orlp

@feersum but for minmax([1, 3], 3) we never reach this

for mid = 1 and [1, 3] we get needed = 2

feersum

@orlp hi is 1 which would give needed == 2 == k.

orlp

for ANY mid < 1 we get needed >= 4

there is never a needed = 3

so when k = 3

muddyfish

16:09

@Mr.Xcoder that's only True

orlp

we can never get needed == k

user202729

@user202729 I think it should be $\lfloor \frac x {lo} \rfloor = \lfloor \frac x {hi} \rfloor$ instead.

feersum

Sorry, I thought k was 2.

orlp

@feersum it was earlier, but then I came up with this counterexample against the needed == k check :)

feersum

Meh, how about the random number solution then?

orlp

16:10

@feersum doesn't really feel like a solution to me ^^

feersum

How not?

user202729

Do you think my solution (the condition) is correct?

orlp

just feels like it's circumventing the issue, I feel like there's a proper solution

Mr. Xcoder

@muddyfish Oh yeah (ಠ_ಠ Mr. Xcoder)

orlp

@user202729 again, that will never happen in the minmax([1, 3], 3) case

at some point we have hi = 1 (which is optimal)

feersum

16:11

Are we running this algorithm on a realistic computer or a magical real number machine?

user202729

@feersum Realistic computer can have infinite precision real.

orlp

@feersum realistic I guess although even better if it was magical

@user202729 and lo will just get exponentially closer to hi, but never reach it

I guess there's a discrete formulation of the same problem

user202729

I know.

orlp

we just write down what values we can reach for each element by spending k splits on them, for all k

user202729

What happens for that case? (I can't run the code)

(is that Python?)

Ok no problem import math resolved it.

But I guess it will terminate? What's the problem?

orlp

16:15

@user202729 you guess?

we need proofs, and not only that it terminates, but also always terminates on the optimal value :)

user202729

I'm not good enough at Python to implement it.

So does it correct for that case? And now you are saying that it is correct and you need a proof, or it's incorrect?

orlp

def minmax(l, k):
    x = l[0]
    if len(l) == 1: return x / (k + 1)
    return min(max(x / (j + 1), minmax(l[1:], k - j)) for j in range(k + 1))

@feersum this always terminates and is correct

@user202729 I'm saying that I'm not convinced it works at all

and just testing it for one case doesn't mean anything

feersum

@orlp But it grows linearly (or more) with k...

user202729

@feersum At least it's provably correct.

orlp

@user202729 the binary search is also provably correct, it just needs time to improve precision

feersum

16:19

I'm just not comprehending the hate for random here.

user202729

You're not convinced that my algorithm works, doesn't means it doesn't work. It just means that I didn't prove it.

orlp

@user202729 can you at least explain why flooring?

I can understand ceil(x/mid)

@feersum The problem reworded is finding the smallest value v such that ceil(x1/v) + ceil(x2/v) + ceil(x3/v) + ... + ceil(xn/v) - n <= k

we can terminate the binary search by finding what the 'breakpoints' of all those ceilings are

and if there is no such breakpoint in (lo, hi) we can terminate

so if for all x ceil(x/lo) = ceil(x/hi) we can terminate, I guess that's what @user202729 meant but I got thrown off by the floor

actually that's not entirely correct

because I said (lo, hi), not [lo, hi]

the bounds are exclusive

and that's the issue with hi = 1 and lo < 1approaching it, because for any integer n ceil(n/1) = n but ceil(n/lo) > n

Luis Mendo

17:45

@Mr.Xcoder It's specially rewarding to reduce code length by abusing a bug :-D

Mr. Xcoder

Isn't it? :D

caird coinheringaahing

18:38

@HyperNeutrino It looks like you may have to save OEIS again :/

HyperNeutrino

19:27

@cairdcoinheringaahing lolWAT noooooo

I don't understand the sequence yet.

It looks semi brute-forceable

Mr. Xcoder

What does rule violations mean (about SE suspensions)?

HyperNeutrino

I just realized why my computer is lagging so much - I'm running multiple copies of Minecraft at the same time ;-;

Mr. Xcoder

I "love" the fact that feature-requests on mother meta are ignored by SE (hoping nobody notices :P)

Erik the Outgolfer

19:44

@Mr.Xcoder lots of se stuff got recently laid off

Mr. Xcoder

@EriktheOutgolfer Really why?

Are they low on money :D?

Erik the Outgolfer

dunno tbf

caird coinheringaahing

@Mr.Xcoder We can't talk about that D:D

@HyperNeutrino Yep, 2 days to go (I think)

Mr. Xcoder

@cairdcoinheringaahing No, I am asking what this means in general. Does it include fraud or just any other violation of the rules?

Erik the Outgolfer

@Mr.Xcoder it's purposefully vague

caird coinheringaahing

19:47

@Mr.Xcoder I think its violations that aren't rude behaviour or vote fraud

Mr. Xcoder

Ok then

Q: If I write a language with overloading rather than vectorization, what should + (kind of addition / concatenation) do on a string and an integer?

caird coinheringaahing

Concatenate

"Test"+5 => "Test5"

Mr. Xcoder

@cairdcoinheringaahing Like "abc"5+ would result in "abc5"?

Lol

Also @cairdcoinheringaahing did you see my message in JHT?

caird coinheringaahing

@Mr.Xcoder Just seen it now :P

Downgoat

20:09

@ASCII-only do you prefer delegates or lambda?

caird coinheringaahing

20:51

@Mr.Xcoder Or add n to the code points. E.g. "abc"+5 => "fgh"

Okx

@cairdcoinheringaahing i think that's too much like vectorization

Transcript for