The Nineteenth Byte

The problem with your problem is that if you have to deal with noisy data, there is no one solution that works for everything, so you sometimes really have to fiddle with the parameters, because no algorithm can magically find out what we as humans perceive as noise and what we consider data:)

Life would be a lot easier if we had such an algorithm.

Well probably not for me, I'd lose my job.

Anush

:)

I don't fully understand why the rbf is the right solution

flawr

9:05 AM

I can assure you, it is not the right solution :D

RBFs are quite flexible in the sense that you can apply them even to very high dimensional data and you can influence the behaviour of your function by the choice of the kernels, and nodes.

And here the neat thing is that you can actually compute the derivative of the smoothed function analytically!

(Because we express the function as a linear combination of differentiable functions, the derivative is again a linear combination of their derivatives.)

New Sandboxed Posts

10:11 AM

0

A: Sandbox for Proposed Challenges

Decode a RISC-V J-type immediate RISC-V is an open processor instruction set, which defines a somewhat typical RISC instruction set. However, in order to make decoding simpler in hardware, the encoding for immediate values tend to be quite complex, with the bits essentially shuffled around. And ...

lolad

10:45 AM

Does anyone have any tips for writing an esolang good for code golf?

Erik the Outgolfer

@lolad wanna join JHT?

Yes please

done

thanks

4:37 PM

@EriktheOutgolfer I thought JHT was just for, you know, Jelly

lirtosiast

5:54 PM

I've been looking for information on golflang design for ages

flawr

@lirtosiast I found this book in our library from 1893

lirtosiast

@flawr oh nice, could you carrier pigeon it to me?

Erik the Outgolfer

@lirtosiast what?

yeah, of course it is :P

oh, the ping was unrelated to the above message, lolad requested access to JHT

lirtosiast

oh

but yeah I really wish there were good resources on the principles of golflang design

I've been discovering all these snippets of knowledge by myself writing Sledgehammer and they feel so incomplete compared to what people who have written complete languages must know

flawr

maybe you should write a book

lirtosiast

6:06 PM

...on what? I just said I don't have most of the knowledge

flawr

but you have probably in the top 0.0001% of the world!

or maybe some zeros more

lirtosiast

I can guarantee that Dennis and recursive know like 10x more

... you know, maybe it's worthwhile to create a wiki or something

What format do resources even go in these days?

feersum

Why? It's not like the world needs more golflangs.

lirtosiast

6:25 PM

Because it's interesting knowledge and should be written down

Some areas of math weren't useful, until they were

John Dvorak

I introduce you to CGCClang. Valid source codes are base64-encoded post IDs referring to CGCC posts. A source code that references a challenge compiles to a program that solves the challenge. Negative IDs (two's complement) reference posts on meta CGCC.

lirtosiast

@JohnDvorak that breaks a standard loophole

John Dvorak

just one of them?

lirtosiast

At least one, which is all that matters

Adám

@lirtosiast Which one?

Erik the Outgolfer

6:33 PM

I totally approve of John's idea; the program does not connect to the internet, and the language is more than just the answer to a specific challenge... :D

lirtosiast

Well, if the compiler postdates the challenge, it breaks that one. If the compiler predates the challenge and it fetches solutions from SE it breaks the external source one.

Adám

@lirtosiast iirc, we voided that rule.

Erik the Outgolfer

the non-competing rule has been abolished :P

Adám

@JohnDvorak Why base-64 and not base-256?

John Dvorak

You're right. It should be base-256.

... or straight up truncated UTF-32 if somebody decides to use character counting.

dzaima

6:36 PM

reason for no golflang guides is that everything is either obviously bad or good for certain things, making any suggestions pointless

lirtosiast

@Adám does the ruling override codegolf.meta.stackexchange.com/a/5772/39328 then?

Adám

@JohnDvorak Since we have more that 256 challenges, and some are solved by one-byters, this lang will still lose some challenges…

John Dvorak

that's inevitable

lirtosiast

I'm still unsatisfied with how we deal with languages postdating the challenge, partly because of exactly this possibility

dzaima

order by score, then do a huffman encoding :p

Adám

6:38 PM

@JohnDvorak We could say that all bytes up to 127 are one-byter encodings for those challenges. 128×265 is plenty.

lirtosiast

aren't there close to 256 challenges won by one-byters?

Adám

@lirtosiast Interesting question.

dzaima

@lirtosiast usually there's a pretty clear line between being specifically implemented for a certain challenge and not imo

John Dvorak

You would expect more than 256 challenges won by one-byters because those one-byters are written in different languages

lirtosiast

Yeah the question is whether people have bothered to write those languages

Adám

6:41 PM

Yea, e.g. many golf langs will have 1-byte increment and decrement built-ins.

Erik the Outgolfer

actually, a byte can be overloaded depending on the number and type of inputs it gets

lirtosiast

@dzaima I agree in practice so far, but what if someone implements a builtin for codegolf.stackexchange.com/questions/176251/the-hungry-mouse that travels between adjacent elements in a 2D array, modifying cells with some function and ending at some ending condition?

Adám

@EriktheOutgolfer Ah yes, there was a controversy about Jelly having implicit ⍨

lirtosiast

It definitely has applications to other challenges, but it's very relevant to that particular challenge.

I understand why we had to kill it, but the non-competing rule was the only fair way to compare golfing languages.

Erik the Outgolfer

for example, the x byte may check for primality for one positive integer input, and for compositeness for one singleton that contains a positive integer, and do more stuff with other kinds of inputs

feersum

6:47 PM

I don't understand why to kill it. I think the non-competing rule is correct.

Erik the Outgolfer

more uses for x: the product of one list of numbers, the dot product of two lists of numbers, repetition of one list by one non-negative integer, in this order, repetition of each element of the list instead in the reverse order, etc.

Adám

@EriktheOutgolfer Sounds like K.

feersum

Also there can be significantly more than 256 challnenges solved by one language, because it can have different behavior depending on whether the input is a number, a string, an array of number; or if it comes on stdin or argv or function arguments, and so on.

Erik the Outgolfer

@Adám surely it's more overloaded than that already :P

feersum

Oh, I guess that's the same thing Erik said in the immediately preceding message, lol.

lirtosiast

6:51 PM

@dzaima There are definitely universal design principles-- consider syntax; assuming your language is a sequence of one-byte tokens, you want a human-writable syntax that allows for most sequences of tokens to have a valid meaning. Therefore, it probably has to be simple to parse, so you can assign those meanings. Prefix, stack-based and trains are all solutions that follow the basic principle. Maybe there are others.

@feersum people were posting answers that just didn't work and claimed they were "non-competing"

dzaima

@lirtosiast ooor you could have not token-based languages - e.g. base encoded (stax), huffman, something more complex

lirtosiast

@dzaima Those are just trivial compressions of a token-based language.

John Dvorak

7:07 PM

Counter-example: bubblegum

lirtosiast

@JohnDvorak okay, not universal, but that principle is definitely applicable to enough golflangs to further develop and write down

dzaima

7:37 PM

@lirtosiast really the only part of token-based that applies universally is that at some point, the language separates the code into smaller things, and i'm pretty sure that falls under "obviously good" because there's literally no other option besides hard-coding every program and it's actions..

lirtosiast

@dzaima sure, but why don't the other elements of that principle apply to most golflangs?

feersum

@lirtosiast This reply is a total non-sequitur to me.

How does that argue for allowing languages created after the challenge to compete?

@dzaima CMC: Design a language that does not separate into smaller things at any level.

dzaima

@feersum chat.stackexchange.com/transcript/message/51021704#51021704, kind of

feersum

@dzaima That separates into smaller things after loading the code and sending it to the appropriate interpretre

John Dvorak

CV CMC: unclear

dzaima

7:46 PM

@feersum well by that look, any language with at least two different programs fails

feersum

Or one non-trivial program :)

dzaima

@feersum does python, but only valid source code is the source of Jelly work?

feersum

No, unless Jelly works.

dzaima

@lirtosiast what other elements?

John Dvorak

Anything decomposes to a sequence of x86 opcodes. Any x86 opcode decomposes to a sequence of micro-opcodes. Any micro-opcode decomposes to a set of lines to be hot

ngn

7:58 PM

and individual electrons passing through transistors

John Dvorak

which decomposes to the step of opening the channel and the step of passing charge through it

feersum

I'm thinking we need to go for something unimplementable

Like iterating a real-valued function.

With irrational numbers and so on.

lirtosiast

8:19 PM

@feersum I misremembered why we changed the rule. I guess that was just an unrelated benefit I noticed.

dzaima

@feersum well, this site is for recreational programming, so why explicitly disallow having fun?

lirtosiast

Anyway, looking at the meta post, it's clear that we want these 4 things:

1) Old challenges are not duplicated (unless they're extremely bad quality by the current site standards)
2) People can submit an answer to a challenge with any language
3) All submissions to a challenge should be competitive
4) All submissions allowed to compete should have an equal playing field

And we seem to have moved from giving up 3 to giving up 4

dzaima

@lirtosiast well the "can't use language made specifically for an exact challenge" directly goes against 3 and helps 4 a lot

New Main Posts

8:45 PM

0

Q: Hello, Game of Life!

Introduction I've been browsing all those hello-world challenges and was thinking "yeah, they're good, but what if we make GoL one?", so here it goes. Challenge Build starting setting for either Conway's Game of Life or other similar cellular automaton (restricted to ones with binary cell stat...

Erik the Outgolfer

9:45 PM

@NewMainPosts hm... I think that one can make an automaton that takes advantage of unique neighborhoods to do something in one step and then bam, the solution is here

New Sandboxed Posts

10:06 PM

0

A: Sandbox for Proposed Challenges

code-challengegame-of-lifecellular-automatahello-world Originally posted on main site, moved here for more suggestions. Better scoring mechanics required. Introduction I've been browsing all those hello-world challenges and was thinking "yeah, they're good, but what if we make GoL one?", so he...

Transcript for