Chat:World/2021-10-12
Coding_Aria: hello world
Thorcode: hi world
derjack: good morning
TheBlueBias: 'nin
Thorcode: good morning derjack
TheBlueBias: :(
Thorcode: ?
derjack: :upside_down:
Thorcode: How can i handle all the number in the last line of fax machine?
TheBlueBias: In the inputs?
Thorcode: yep
Thorcode: at the heart there is alot of input and It make me dizzy
TheBlueBias: What language do you use?
Thorcode: Python
TheBlueBias: Well something like: [int(i) for i in input().split()]
TheBlueBias: But that's probably already in the base code.
Thorcode: nope
Thorcode: it just input()
TheBlueBias: Oh. Right.
Thorcode: and in the fifth test case the input so long :(
TheBlueBias: So. I gave this piece of code to you, but you'll have to learn about it. Search "list comprehension"
Thorcode: thanks
Thorcode: there is alot of stuff I didn't know in python :(
TheBlueBias: Do you follow any course?
Thorcode: yep
Thorcode: a course on w3school
TheBlueBias: Well you'll probably learn what's missing soon.
Thorcode: list comprehension doesn't help me
Thorcode: oh I get's it
Thorcode: thanks TheBlueBlas
TheBlueBias: :thumbsup:
Thorcode: alot
Thorcode: Wontonimo have a nice puzzle
TechnoBot: Howdy'
TechnoBot: Hello?
Uljahn: :door:
Thorcode: hi
TheBlueBias: Bonus day on ICPC! \o/
struct: why?
Seii: iza no malagasy ato
struct: English in this chat please
Seii: sorry, can't write english very well
AntiSquid: there is #jp
AntiSquid: although what you wrote doesn't look like a spoken language at all
eulerscheZahl: TheBlueBias my reaction was more like :rage:
struct: I agree euler
struct: its unfair
eulerscheZahl: no I'll ask my boss for an extra day off. still pissed, tell me earlier and I can plan. and be less disappointed when I drop in the last hours
AntiSquid: ah malaysian ? seii, well dont think anyone else here speaks it
struct: Did they give any reason on why they extended the contest?
Seii: sorry everyone
AntiSquid: you should be
derjack: oO
AntiSquid: it was a joke, sir bubbleeyes
eulerscheZahl: no reason. just "The contest duration has been extended to 5 days. The actual finish time is October 14, 2021, 00:00 UTC**"
eulerscheZahl: the video announcement always said 5 days. maybe CF did it wrong in the first place?
eulerscheZahl: and now CG down?
eulerscheZahl: CF*
eulerscheZahl: Oops! Probably Codeforces can't be reached right now or your Internet connection is broken. You can try:
eulerscheZahl: https://codeforces.com/
AntiSquid: ya
AntiSquid: it's down for me too
eulerscheZahl: sounds serious. why not extend it to a full month because of this?
AntiSquid: seen some other contests get randomly extended last minute, makes me wonder is it a "we don't like who's at the top so let's give it another try" sort of thing :D
eulerscheZahl: i predicted tourist, NVAL, T1024 and wleite at the top ranks. only tourist still missing
eulerscheZahl: but last time he also appeared close to the end already. has already submitted but really behind in ranking for now
struct: nval comented on the day added
eulerscheZahl: he's not the only one who ranted
struct: I only saw 2/3 comments
struct: but was on the official contest page
eulerscheZahl: mine was the 4th comment about it
eulerscheZahl: thought the more users write about it, the more likely they see what they did
AntiSquid: i would invite some coworkers to post a rant too if i was angry about it
AntiSquid: 502 Bad Gateway on all pages
TheBlueBias: Yeah. I understand euler. But since I didn't had the time to do a lot, for me it's more a good news...
TheBlueBias: You're participating to ICPC too Squidy?
AntiSquid: who's this guy calling me squidy ?
AntiSquid: no, i looked it though
eulerscheZahl: i would never do that. I always write squiddy with 2 d
TheBlueBias: Meh... He doesn't recognize me... :(
Overbed: where do i learn python 3 for free
C26_1: you can googles for yourself
C26_1: or searching in Youtube
Mundgus: w3scool
Mundgus: geeksforgeeks
Mundgus: for python learning
C26_1: @Mundgus, yes explanatory websites
TheBlueBias: Overbed https://www.letmegooglethat.com/?q=learn+python+basics
TheBlueBias: :3
C26_1: I've never seen a website that's like this
C26_1: Set = contains unique things
Overbed: is hyperskill.org a good website to learn python 3 from?
Overbed: it isnt free but my school can fund me a bit
C26_1: Array/List = contains a lot of things (may including copied things)
Mundgus: If you search paid website, use udemy
Mundgus: udemy.com
C26_1: @Mundgus, struct (Only in C/C++/C#) = short-words for structure, somethings than contained multiple information about a object
C26_1: @Mundgus, dictionary = one-to-one correspondence of info-checking
C26_1: does that fact true or just simply false
Mundgus: struct can contain functions, not just variables
derjack: in c++ struct is just a class without private stuff
Mundgus: Yep
struct: I think it can still be private
struct: its just public by default
eulerscheZahl: do you have inheritance for structs?!
eulerscheZahl: i always thought that's the main difference
eulerscheZahl: and for C# structs are value types and not object types
struct: you do have
eulerscheZahl: ok. big difference between C++ and C#. to no surprise
struct: They are exactly the same in c++, the only diference is that struct is public by default and class is private
derjack: trust this guy, he knows a thing or two about structs
derjack: loa has no longer WIP in title, but its still WIP?
eulerscheZahl: no, it's VIP now
derjack: :tada:
miszu: finished 22nd in cvz. Need. To. Optimize. More
Uljahn: :tada:
miszu: what reasonable algorithm can be used when you have (2000x2000)^30 search space?
eulerscheZahl: got the extra day off. still not happy about codeforces
eulerscheZahl: just bruteforce it miszu
eulerscheZahl: you are exaggerating the search space, it's only (2000x2000 x PI/4)^30
Uljahn: NN or search space discretization i guess
miszu: eulerscheZahl akchually.. it's (PI*1000^2)^30
eulerscheZahl: >>> 1000**2 * math.pi == 2000**2 * math.pi/4 True
eulerscheZahl: even python with possible rounding issues agrees that they are the same
miszu: my trigs sucks then lol. I didn't get how you added a pi/4 at the end
Uljahn: what's possible rounding issues in python? :thinking:
eulerscheZahl: >>> sum([0.1]*10) 0.9999999999999999
eulerscheZahl: case closed
Uljahn: is it python's issue though?
eulerscheZahl: IEEE754
derjack: 0.1 + 0.2 = ???????
miszu: how good CMA-ES is?
eulerscheZahl: how good your English is?
miszu: anyone here used it and was satisfied?
miszu: mmmm... English is 3rd language so not the best :(
eulerscheZahl: sorry, just teasing ;)
miszu: haha that's cool :P
eulerscheZahl: for me it's the 2nd language. I surely write some wrong or even stupid stuff as well
miszu: I want to an english college to learn english properly. Man I struggled because you need to pass an english "exit exam" which is like an 2k word essay about ideas from the author
miszu: went*
jacek: huh? i thought there was english in canada
miszu: if you live in Quebec it's a bit complicated
miszu: they don't like English much
jacek: who likes them :v
chungangli: Hello, everybody, who can help me?
chungangli: I want to change the account email, but I can't do it.
chungangli: I am using gmail for this account, but I want to use outlook email for it. but I have no way to do it.
DomiKo: miszu I used it for some tasks, and it was pretty good, BUT it really depends
miszu: DomiKo do you have a good reference that explains the algo clearly? All I can find is very mathy equations and I can't make sense of it
miszu: I understand the high level but all the sigma etc. assumes that the reader knows exactly what it means
DomiKo: algo isn't really easy, so I don't know if you can find any explanation without math
miszu: at least they explain the pseudo code the operation
miszu: if not I will continue to search
DomiKo: https://blog.otoro.net/2017/10/29/visual-evolution-strategies/
DomiKo: there you can find some visual effects
DomiKo: that could help
chungangli: Dear DomiKo, could you help me?
DomiKo: about mail change?
chungangli: yes
miszu: thanks DomiKo
DomiKo: right corner, there is a arrow, click it, then go to settings
DomiKo: and in setting in "Login information" you can change email
StarsAreFire: For the https://www.codingame.com/ide/puzzle/aneo Is the distance stated - An integer distance representing the distance of the traffic light from the starting point (in meters). From the very start of the drive or from when you enter the zone?
chungangli: I tried, but it is needed to enter password, but I have no password becasue I am using google login now.
Uljahn: those are the same i guess
Uljahn: StarsAreFire
DomiKo: if you are using google login, you need to get a password
TheBlueBias: chungangli use the "forgot password" option to set a proper password, then you'll be able to change your email
DomiKo: do what TheBlueBias told you
LeeDeffo: i have no idea what im doing right now
TheBlueBias: Well. I have no idea of what you're doing right now, if that rassures you.
chungangli: Thank you, friends, it works well. :)
DomiKo: nice
TheBlueBias: :tada:
TimothyAlexisVass: Sometimes when people reverse their thinking, they move sideways without looking at the backside underneath their own position.
jacek: ~paulo coelho
miszu: DomiKo lol in the page: "CMA-ES is my algorithm of choice when the search space is less than a thousand parameters. I found it still usable up to ~ 10K parameters if I’m willing to be patient." cvz: *laughs with the billions of search space*
Uljahn: paramters not states
HalfTangible: We're supposed to share our code, right?
Uljahn: no, this is optional
miszu: my bad
miszu: is it possible that CMAES gets stuck in a local optimum given enough time?
DomiKo: enough time is a weird sentence
miszu: enough generations
DomiKo: I'm not 100% sure, but I would say yes.
DomiKo: but with big enough population, it would be really rare
miszu: I will try to give a shot with cvz. At least the simple one not with the integrals and all stuff
jacek: :scream:
miszu: all I need to figure out is how to generate random points from gaussian distribution and how to apply when you want to make a series of moves
miszu: linking mean and std with many moves ahead
miszu: TIME TO CODE CMA-ES!
TheBlueBias: Don't scream.
jacek: madness
TheBlueBias: https://www.youtube.com/watch?v=KwIe_sjKeAY
miszu: jacek you think it's overkill to use that algo?
miszu: glad c++ has a normal_distribution class in std. One problem less
jacek: its alright
TheBlueBias: don't think twice
jacek: but cma-es sounds similar to madness
darkhorse64: Since people started encoding NN brains in UTF16 for CG, sky is the limit, miszu. (100k too)
miszu: cmaes is not so bad. It's just computing a normal distribution over X% best sample
darkhorse64: :popcorn:
nikogabro: hello can anyone tell me whats catch in codingame sponsored contest??
TheBlueBias: No. All the interest is to figure it out by yourself.
nikogabro: okay whats your best?
Andrew_Clark: I think I did well
nikogabro: like how much?
nikogabro: 207?
TheBlueBias: You're not speaking of the same thing...
nikogabro: xd
jacek: oO
miszu: if my parameters are x and y but I simulate 10 moves ahead does that mean I have 20 parameters?
Uljahn: i think yes
Wontonimo: are you talking about a GA miszu ?
Wontonimo: if so, that is how some people implement a GA
Wontonimo: an implementation detail is that on the next game round, you move all the parameters 1 position.
Wontonimo: also, some simulations work better by keeping dx and dy instead of x and y
Uljahn: also you can use complex numbers :smirk:
Wontonimo: oh? tell me more about that
Uljahn: i mean one complex number instead of (x, y) pair
miszu: Wontonimo talking about CMAES
miszu: just understanding the math
miszu: I use dx and dy
miszu: I understahnd you compute mean and sigma for each parameter but I don't get it how the sigma(xy) comes into place
Wontonimo: what is CMAES?
Uljahn: CMA-ES
Uljahn: evolution strategy
Wontonimo: oh, cool. i just learnt something. Thanks
ZarthaxX: guys any recommendation on a cpu?
ZarthaxX: like intel or amd for instance, or a specific model
miszu: amd ryzen are good
miszu: the newest generation\
ipsissimus: ^
ipsissimus: note however, that if you get an amd cpu, youll need a gpu as well, as most amd chips dont have integrated basic graphics like intel
ipsissimus: unless thats changed recently
ipsissimus: *shrug
ZarthaxX: no problem about that
struct: price range ZarthaxX?
jacek: integrated gpu is crap
ZarthaxX: have a gpu, just wanted to know what cpu is good for programming purposes and i guess a bit of gaming
ZarthaxX: struct ehhh hard to tell with the coin conversion hehe
ZarthaxX: i was looking at i7 11gen for exmaple that is 330$
ZarthaxX: accesible here
ipsissimus: no more than a 4 core id say. no need to go crazy with threadripper or anything if you not trying to do some blender rendering etc
ZarthaxX: okey
jacek: but if you ml stuff, the more cores the merrier
ipsissimus: yeah
ZarthaxX: that's something i was worried about
ZarthaxX: if i plan to do data science stuff
ZarthaxX: ML etc
ZarthaxX: would be nice to have 8 cores 16 threads?
ipsissimus: sure, if youre looking for max cores though, i think amd is best bang for buck at the moment
jacek: thats why i have amd. 16 threads
ipsissimus: intel been slacking for like 5 years now
struct: the 11700k has avx512
ipsissimus: thats a over clockable chip i believe
struct: you can even go for the kf
struct: which doesnt have the integrated gpu
ipsissimus: yeah dont pay for stuff you dont need
ZarthaxX: well in argentina the version without F is cheaper than the one with F haha
ZarthaxX: but as you are saying, intel has been kind of bad in last few years, but i didnt know if the i7 11gen was different
ZarthaxX: or the i7 10gen
ipsissimus: i have a great basic system ryzen 3600X and rx 580. does the job and gives me 1080p 60fps which im happy with.
struct: I think the 11th gen is fine
ipsissimus: was around 800 bucks total build
ZarthaxX: avx512 thingy is interesting but i think useless struct haha
ZarthaxX: tempting tho :(
ZarthaxX: nice ipsissimus
struct: Whats the 10850k / kf price there?
struct: its 10th gen though
struct: but 10 cores
ZarthaxX: an equal option in terms of money for i7 11 gen was the ryzen 7 3700x
ipsissimus: hyperthreaded?
ZarthaxX: mmm lemme checj
struct: yes ipsissimus
ipsissimus: i think x variants of ryzen chips are worth it for your use cases
ipsissimus: ryzen just so good for an chip class
ipsissimus: amd went hamm
struct: it depends on the country you live in mostly
ZarthaxX: i9 10850k is like 1/5 more than the i7 11700
struct: On some countries theres barely any difference in price
struct: ah ok
ZarthaxX: ipsissimus you think so? okey..
ipsissimus: ye but in US i think intel overpriced
ZarthaxX: i have seen reviews everywhere recommending amd tbh
ZarthaxX: but i didnt know if anything changed now with this 11gen
ipsissimus: or maybe ryzen underpriced idk
ZarthaxX: ryzen here seems to be above intel in terms of price now
struct: yeap thats what I mean
ipsissimus: ye that demand for a superior product :O
ZarthaxX: ryzen 7 3700x is equal in price to i7 11700, has same cores and threads
ipsissimus: but i think nows a bad time for chip buying no?
struct: but 2nd gen
ipsissimus: what with semiconductor shortage
ipsissimus: ?
ZarthaxX: according to this page tho i7 11gen is better https://cpu.userbenchmark.com/Compare/Intel-Core-i7-11700-vs-AMD-Ryzen-7-3700X/m1450761vs4043
struct: also new cpus coming soon I think
ZarthaxX: but that page says the i7 11gen is better and later on every page says ryzen is better
ZarthaxX: :(
miszu: ryzen 5900 is very powerful
struct: I doubt the 2nd gen ryzen is better than the intel 11th
miszu: that's the one I will buy once I have an rtx
struct: Maybe in some scenarios
ZarthaxX: uh the 5900 one seems very expensive :8
ZarthaxX: :(
ipsissimus: yeah intel chips arent bad we should be clear. just USUALLY more expensive. and sometimes expensive enough that you can get a better chip from amd for less
ZarthaxX: struct its hard to tell without testing it :sweat:
ZarthaxX: right
ipsissimus: doesnt intel have ARC for benchmark testing?
ZarthaxX: it makes me doubt as i have seen that every page recommends amd because its better at multicore
ZarthaxX: so i dont know if even having 2 cpus with same cores and threads, and a benchmark saying intel one is better, i still shouldnt believe it haha
Babali: Bonjour
MmODeadlySins: Bonjour ;)
ZarthaxX: hey
MmODeadlySins: You'are Welcome :)
ipsissimus: i dont think you can go wrong with ryzen at the moment. unless next gen comes out with crazy new tech
ZarthaxX: okey
struct: hi
TheBlueBias: :french_bread:
ipsissimus: but it may not work out price wise dependent on location
Babali: @MmoDeadlySins ça va gro ?
Astrobytes: #fr guys
TheBlueBias: Baguette
ZarthaxX: ipsissimus thing is i have these 2 cpus at same price, and i didnt know what to pick
ipsissimus: which 2?
ZarthaxX: i mean these https://cpu.userbenchmark.com/Compare/Intel-Core-i7-11700-vs-AMD-Ryzen-7-3700X/m1450761vs4043
Babali: Berret
MmODeadlySins: @babali lol
WisTala: English here thanks
ZarthaxX: and the i7 10700 is a 1/5 cheaper option too for those, dont know if worth
ZarthaxX: https://cpu.userbenchmark.com/Compare/Intel-Core-i7-10700-vs-AMD-Ryzen-7-3700X/4077vs4043
ipsissimus: ryzen 15$ cheaper
ipsissimus: ez choice ;)
ZarthaxX: ignore that
ZarthaxX: here is same price
ZarthaxX: i just wanted to show models and benchmarks, not prices as they dont make any sense when talking in different countries :P
struct: ah here the 3700x is way cheaper
struct: 270 vs 320
ZarthaxX: damn :(
ZarthaxX: maybe they are equal here because people is buying ryzen mostly
ipsissimus: well that says something too no?
ZarthaxX: what thing?
ipsissimus: that everyone buying them up and its driving up the price
ZarthaxX: ah
ipsissimus: and its not like amd spamming marketing like intel does
ZarthaxX: yeah its true
ZarthaxX: well in the benchmark page, the amount of benchmarks for ryzen one is huge
ipsissimus: really though, disappointed in intel. they need to step it up. Competition is good for consumers
ZarthaxX: like 10x and more times intel one
ZarthaxX: yeah :(
ZarthaxX: so, ryzen option is good for programming too?
ZarthaxX: i just dont know how much multicore benefits here
ipsissimus: yeah like I said, ryzen is just a sick with the (zen??) architecture
ZarthaxX: aight
ipsissimus: they really went hamm on this product skew i think
ipsissimus: seen alot of true believers come out because of ryzen
ZarthaxX: good to know
ipsissimus: yeah most people dont know but the only reason amd was in the game with intel owning like 80% market share in the past was that amd had the 64 bit architecture patent, and they licensed it to intel. but with recent products, their market share has soared
ipsissimus: take a look and see for yourself
ZarthaxX: oh nice, will check it out ty :)
ipsissimus: https://hexus.net/business/news/components/147224-amd-briefly-passes-intel-desktop-cpu-market-share/
ZarthaxX: annoying to be in this moment of the year where there is a 50 50 tho haha
ZarthaxX: and with these 11 gen from intel that dont have much time on the market to have a clear judge about them
ipsissimus: yeah but both manufacturers going to have a hard time with supply. so prices going to get raised
ipsissimus: prophecy detected
ZarthaxX: probably haha
struct: I dont know how well 12th gen will do
struct: with multiple diferent cores
struct: like 16 cores but 24 threads
struct: only 8 cores have 2 threads
ipsissimus: likely will be OS dependent i think right? some OSs handle core abstraction better than others
struct: no idea
ipsissimus: i have the other extreme than zarthaxx, i like to do small projects on toasters. really makes you figure out how to be efficient
jacek: like how to fit NN in 100k limit
ipsissimus: ye. tho im trash at programming, i know hardware better cuz im EE. but thats why im here :) to get trashed by my competition and look at their code
TheBlueBias: Toasters are less expensive though.
ipsissimus: thats right. and if you can get it to work on a potato. itll work beautifully on company high end employee issued comps
matez: The clash of code page sometimes does not work: when the clash would start it does not give the coding page and you cannot do anything just became the last player :(
TheBlueBias: Just refresh the page and you'll get the IDE
matez: I will test it...
eulerscheZahl: https://codeforces.com/blog/entry/94906?#comment-850177 reasoning for extending the Huawei contest got downvoted so much that it's hidden by default :D
Astrobytes: " So was it just a bad communication (pun intended) " :D
eulerscheZahl: I couldn't resist to do that edit on my post :P
eulerscheZahl: i'm getting wrong answers now and I can't find my bug
eulerscheZahl: works fine with offline tests. I have no idea how the server cases look like
struct: Do I need to give a solution to every flow?
eulerscheZahl: no
eulerscheZahl: if you do, it would be a perfect score
struct: dont think it would be perfect
struct: Because I could not take the best path
struct: Unless I understood the problem wrong
eulerscheZahl: it's about how many valid paths you can create
eulerscheZahl: the distance is the tie breaking in case of same number of paths. looking at the scoreboard you can completely ignore it
struct: oh I see
struct: thanks
Rich_Engler: When your binary search is one step too slo w><
Rich_Engler: For the most part it works, unless the value being looked for is at one of the extremes...
jacek: batman eh
Rich_Engler: yep
Rich_Engler: 85%
Rich_Engler: I just run out of time in test 6
os-moussao: Test 2 didn't make sense for me, if the ball in (0, 0) went the the hole in (2, 1) it would be already 3 moves, yet the ball can is only make 2 moves????!!!!!!!
os-moussao: what do you think guys?
struct: os-moussao which puzzle?
derzinek: Hello guys. I'm new here and I need some help. Trying to solve chuck norris. Can't pass third test. ASCII for % char is 37 and binary for 37 is 100101. When I encode it result is: 0 0 00 00 0 0 00 0 0 0. My code works correctly but the test says it's 00 0 0 0 00 00 0 0 00 0 0 0. Why?
struct: Needs to be 7 bit binary
struct: so 100101
struct: would become 0100101
derzinek: I think I missed that part. Thank you so much struct.
struct: it needs leading 0's
struct: np
struct: leading or trailing
struct: i cant remember the right word
derzinek: "i cant remember the right word" never mind, I'm not so good at English :))
struct: basicly you need 0's on the left until you have 7 bits
struct: :p
jacek: أوه حقا
jacek: oh really
jacek: Automaton2000 what do you think of this
Automaton2000: i make a new one
Rich_Engler: Trying to solve batman ep1. I don't really see how I can optimize the code further. All tasts pass except for 6 where I am one move too slow
struct: seems like an infinite loop
struct: batman doesnt really require you to optimize the code
struct: ah
struct: one move too slow
struct: sorry
struct: I read it wrong
lolimagine: I GOT
lolimagine: unbanned!!!
Kitis: lol
Astrobytes: let's see how long that lasts . . .
Chrisidris21: rookie lua coder
jacek: beautiful https://i.imgur.com/kH0oSAK.png
2ReGul4r: Hey can someone explain me how I can approve contributions? I got 50+ games of clash of code, but I still cant do it :(
jacek: shouldnt it be more?
2ReGul4r: it says: Any player who has played more than 50 clashes has approver rights.
jacek: so you have 50 or 51 games played
2ReGul4r: i got 51
jacek: maybe it takes time to acknowledge, i.e. the ranking calculation is done once per day
struct: search racer jacek?
2ReGul4r: yeah thats the only tought I had why it would work
jacek: struct eeyup
jacek: i though id try nn there before trying in csb
struct: nice
struct: how many different moves do you allow?
jacek: for GA every move
jacek: -18..18 and 0..200
struct: but this wont be for the NN right?
struct: wont be the same
struct:
Westicles: only 49 clashes
Westicles: you have to look on the leaderboard
jacek: yeah, id probably quantify for starters
jacek: at first im gonna do supervised based on moves from GA
Astrobytes: Quantize you mean?
jacek: quack
ja_fica: I'm having an hard time to convert an array of 128*128 float values to fit in 100k size code
Astrobytes: I was asking seriously
struct: ja_fica do you use 32 bit float?
ja_fica: I do, but I can discard the last 16 bits
ja_fica: and the remaining of the mantissa
ja_fica: *and keep
struct: if its 16 you can use special chars
struct: I think
struct: the cg editor counts chars not bytes
struct: so you can use 4 bytes chars
MSmits: some chars are counted as 2 chars.
ja_fica: the problem is that \0 and \n characters exist
MSmits: (or something)
MSmits: ja_fica let me check how i do it.
struct: I never did this before so I dont really know the details
ja_fica: im having trouble in \0 and \n
MSmits: ja_fica I skip a few characters
MSmits: http://chat.codingame.com/pastebin/233c8046-51fc-4c1c-91b6-feb96730ec86
MSmits: I start at 'a' + 95 to skip most of the control characters
MSmits: and then there are 2 more of those in the 8k range, i also skip
MSmits: stay below 50k or so and you should be safe
MSmits: I only use 14 bit from a unicode character myself, but 15 bit should be safe
MSmits: it's just a bit annoying decoding it, because you have to be careful about how to do the -2 for results higher than 8232
jacek: welp i use 1 weight - 1 character
MSmits: yeah but how do you avoid those control characters?
MSmits: have you just been lucky not hitting them accidentally?
jacek: lower resolution
MSmits: dont know what that means
jacek: + its somewhat homemade ugly solution
MSmits: ok so you skip the bad characters?
MSmits: like this one:
jacek: http://chat.codingame.com/pastebin/c757cb95-b960-477c-b57e-7e3c79517ac7
MSmits: https://www.codetable.net/decimal/8232
MSmits: mmh not sure i understand that completely, but are you just skipping way more characters than you need to?
jacek: probably. here the 'resolution' is 1/3000.0, enough for most purposes
MSmits: yeah but i am not worried about the accuracy of it
MSmits: i am worried about it converting to unicode 8232 which is a line separator
MSmits: thats what ja_fica was talking about
ja_fica: Too complex for me to get it right was
ja_fica: *way
jacek: unicode 8232?
MSmits: ja_fica it's not too hard, if you start at a + 95 and skip 8232 and 8233
jacek: i just care it will not exceed 50000 something
MSmits: this one jacek: https://www.codetable.net/decimal/8232
MSmits: it makes your string go to the next line when you paste it
ja_fica: yes
jacek: i never encountered it then
MSmits: i only did recently
MSmits: (my book got big)
jacek: and the visualiser is done somewhat :tada: https://streamable.com/37v53b
MSmits: that looks cool
ja_fica: MSmits, with your code, I cant print the characters
ja_fica: they get implicit converted to %255
MSmits: no, the function is bigger, but it's C# code, so i dont know if it'll help you
MSmits: this is just the part that skips the bad unicode chars
MSmits: also i dont do floats (yet)
ja_fica: I can do C# as well
MSmits: ok i'll share the code, but its complicated
ja_fica: Well Im not worried about floats
ja_fica: I can use union to integer
MSmits: http://chat.codingame.com/pastebin/9ff1a15c-a645-42be-b495-5d98e582d263
ja_fica: thanks, I'm having this problem at a long time ago
MSmits: I use 6 bits per yavalath move
MSmits: since a character holds 14 bits, the moves are sort of split between different unicode chars
MSmits: the bitindex is where in the current number I am at
MSmits: and the bitnumber is the number for that character
MSmits: this will eventually become a unicode char
ja_fica: thats genius
ja_fica: Thats exacly what I tried to do
MSmits: thanks, i use different systems for different games
MSmits: in bandas i put 10 moves in 1 char =)
MSmits: 10 solved states i mean
ja_fica: but can't we make 14 bits of info into a single char?
MSmits: yeah
MSmits: so a solved state is like 1.5 bits :)
MSmits: W/L/D
MSmits: thats less than 2 bits
MSmits: I do a ternary representation here
MSmits: 3^10 fits in 1 unicode char
MSmits: so you can fit 10 WLDs in 1 char
MSmits: but for Yavalath it's just 2 moves and a little extra, because it's an opening book
ja_fica: Im a litle confused how to use it for 14 bits but I'm definitely use C#, thanks a lot
MSmits: it's easy. First move fits in bit 0-5 of first char, Second move fits in 6-11 of first char, Third move fits in 12-13 of first char and 0-3 of second
MSmits: thats if you use 14 bit, but i could have used 15
MSmits: it's just messy to code, you're going to make bugs. But the concept is easy
MSmits: that list of integers i feed in, all has integers that fit in 6 bits ofc (yavalath has 61 hexes)
Chrisidris21: first time learning lua
struct: I see that you got 1st again
struct: nice
MSmits: Thanks, I remade the book, it's much bigger now
MSmits: and the bot is new too
MSmits: better TT and pruning
struct: you are doing c4 atm right?
ja_fica: I finnaly got around flatten neural network weigths. Thats my final step to copy weigths. Then all that I'm missing is the training process. Alpha-zero like code its hard to transpose to C++
MSmits: remaking C4 with the same kind of TT. Locally will redo also for meta mcts, I have a great idea for pruning. Maybe i can solve C4 with that
jacek: solve nine men morris
ja_fica: 6x7 is solved in 5seconds
MSmits: I was thinking about the fact that we use only 3 possible options out of 4 with our bitboard representation: Namely empty, p1 played or p2 played. There's not an option for occupying both p1 and p2 bitboards
ja_fica: you can use transposition table with 72 bits
MSmits: so what if I use that for representing parts of the board that can no longer affect the outcome?
MSmits: like cells that can no longer make 4 in a row
MSmits: it does not matter if they are blue or red
ja_fica: static already played positions?
MSmits: yes
MSmits: i can make many many transpositions with that
MSmits: reducing the statespace by a lot
MSmits: so when creating a new state locally, I check the whole board, all useless cells are coloured as this other color
ja_fica: thats happens if you go full minimax
ja_fica: and not MCTS
MSmits: and represented by both p1 and p2 filled
MSmits: this is just for local meta mcts though
MSmits: it will be too slow otherwise
MSmits: drawn or nearly drawn games only solve around 17 or 18 plies before the end. I think i can solve these much sooner with this TT trick
MSmits: or at least re-using the results much more
MSmits: btw, i am not interested in using this to retain/regain nr 1 C4. I could just do that by making a bigger book. I like a shot at solving it
struct: I think ill code my mcts to change the value of the node instead of having the status var to check if its solved
struct: Just to try the avx select again
wizaral: Hello. How to reopen school/job chat if I've closed it?
MSmits: ohh right, i'd be interested to see if you can make it work struct
MSmits: no idea wizaral
struct: i have a version but it doesnt perform better
MSmits: yeah i know, i meant if you make it work better
struct: mainly due to the fact that I need to do some weird stuff to the status var
wizaral: (╯°□°)╯︵ ┻━┻
MSmits: struct you could try it with a game that has no draws
struct: wizaral was it a chatroom?
MSmits: then you can avoid status altogether
MSmits: (just set extreme negative wins or positive wins for solved games)
struct: I still think it wont do much of a difference
wizaral: I used join command and it reopened
MSmits: probably not, but that won't stop you
struct: true, ill try to do something on ICPC first though
MSmits: gz wizaral :)
MSmits: what is this ICPC
struct: https://codeforces.com/contest/1576/problem/A
MSmits: ohh ok, looks cool
MSmits: jacek i was going to solve nine men's morris at some point. Is it up yet?
jacek: its in contributions
MSmits: yeah i know, been there for a long time
struct: I tried playing it vs ludii
MSmits: it's solvable, but requires some trickery
struct: it started smiling at me before I placed all the pieces
struct: :/
MSmits: haha
MSmits: it's exactly the kind of game i like to try and solve so I hope it gets out of contribution soon.
MSmits: you can't fit the whole solved result in a bot as a book, so it's not like you can use it very well ingame
MSmits: but it would be cool to try and train a NN on the solved results
Astrobytes: weren't you going to approve 9mm jacek?
Astrobytes: it only needs one more since eric approved earlier
**Astrobytes killed the chat
TheBlueBias: With a 9mm...
Astrobytes: lol
Astrobytes: that was 9 Mens Morris btw in case anyone didn't catch that
MSmits: naw, we just assumed jacek visits people who submit solved games with a 9mm
MSmits: that jacek, always trolling people
Astrobytes: Hey, Poland isn't *that* wild lol
MSmits: I'm sure Russians think it's the wild west
Astrobytes: I'm not convinced
MSmits: they have wild west themeparks at least :P
Astrobytes: lol, I don't want to know :D
Astrobytes: The image conjured up in the mind upon thinking of that is quite enough
MSmits: hehe ok
Astrobytes: MSmits: Why did Erwin Schrödinger, Paul Dirac and Wolfgang Pauli work in very small garages?
Astrobytes: They were quantum mechanics!
Astrobytes: (before you look it up)
MSmits: oh nice one :)
Astrobytes: A blowfly goes into a bar and asks: "Is that stool taken?"
Astrobytes: (blowfly larvae eat sht)
Astrobytes: Done for today, I'm here all week
MSmits: lol
Astrobytes: lo
Astrobytes: No no, this version has "William Tell has stretched his bow till it won't stretch No furthermore and/or it may require a change that hasn't come before." after a few rearranged lyrics. I mean they forgot shit all the time lol
Astrobytes: ahhh ffs
Astrobytes: So wish there was a remove message button
Astrobytes: (wrong tab, please ignore the above)
miszu: man get sample from multivariate gaussian is so complicated
Astrobytes: hahaha, fun with stats
miszu: I thought you can plug and play but NOOOO... you have to find a matric A such that A*A^t = Sigma
Astrobytes: It's not a straightforward algorithm
Astrobytes: I mean, I haven't used it, but I've checked it out (the maths at least) and I would make sure you understand the maths first before attempting the algorithm itself
Astrobytes: Are you sure this is a good idea for CvZ?
miszu: cvz would work in theory
miszu: because cma es is good for problems that the reward is much later
miszu: such as cvz
miszu: there are scenarios where you score big points just at the end
Astrobytes: Hm, I get your reasoning
Astrobytes: And good luck with it! Nice placement with the GA btw, you don't just want to fine tune that for a better score?
miszu: I did flat MC
Astrobytes: Aha!
Astrobytes: Told you it was decent
MSmits: Sent you a reply miszu. Going to sleep now
MSmits: gn all !
Astrobytes: gn MSmits
miszu: gn MSmits
miszu: I did flat MC before but I did a mistake
miszu: I chose number between -1000 to 1000
miszu: however rarely you get 1000
miszu: so I fixed thanks to struct and increased to like 2000 so like 50% chance you get 1000
Astrobytes: that's what I meant before about ranges
Astrobytes: Have a higher+lower range than you need and clamp the result to your legal range
Astrobytes: well, not clamp, but take the appropriate value
Astrobytes: You won't go back to trying your RHEA?
miszu: I was too stupid to understand what you meant before :P
miszu: RHEA?
miszu: what's that?
Astrobytes: lol Ah, rolling horizon evolutionary algorithm
Astrobytes: The GA type you were using previously
Astrobytes: GA/EA
miszu: GA I tried but gave me suboptimal results
miszu: but maybe I implemented wrongly
miszu: the crossover part
Astrobytes: That's in its nature but it should be better than flat MC
miszu: I simply took the delta X and delta Y from different parents
Astrobytes: this is a conversation for later for me I'm afraid, I'm going to bed. I think it's worth giving it a go again
miszu: i will
miszu: thanks!
miszu: but need some research about crossover
Astrobytes: there's a lot of different ways, start simple and experiment
Astrobytes: Anyway, I'm out, see ya later :)
miszu: take care!
Thorcode: hi
Wontonimo: hello all
Wontonimo: hey khalis
Wontonimo: what brings you here?
Wontonimo: not much eh?
Hyperplane: wtw ;]
Wontonimo: hey Hyperplane
Wontonimo: what are you working on?
Westicles: mom called them both to take out the garbage
Wontonimo: it happens
OofButBetter: hello. i am new to this website!
Wontonimo: hey
Wontonimo: how did you find out about it?
Mauprogamer333: can you recommend me a book of algorithms?
OofButBetter: my tutor showed me it
Wontonimo: oh. what have you seen so far?
OofButBetter: only a couple of taks, but i am having lots of fun!
Wontonimo: there is this python cookbook http://bedford-computing.co.uk/learning/wp-content/uploads/2015/10/Python-Cookbook-3rd-Edition.pdf
Wontonimo: what programming language are you interested in
Wontonimo: oh, that link i just sent could be good
OofButBetter: i also write in c++, but mostly python
Wontonimo: and also there is this site to practice some easier puzzles in python https://www.hackinscience.org/exercises/
Wontonimo: once you have the basics like loops, arrays, functions then try some of the puzzles here
Westicles: after you do 50 clashes you can approve my cryptic reverse contribs
Wontonimo: no no plz don't
Wontonimo: oh, i read that wrong
Wontonimo: i thought you said "approve any cryptic reverse contribs"
Wontonimo: please do approve Westicles' reverse contribs
Westicles: heh heh, well no you had it right at first. I watched twitch and they mostly hate them
Wontonimo: my secret is that i don't watch twitch, so I don't know how much people hate my contribs
Westicles: binarysearch seems to be the big thing these days
Wontonimo: what kinds of things have you done with python and c++ so far OofButBetter
OofButBetter: uhh i programmed on those languages and learend to use them but nothing else really
Wontonimo: this is one of the 'intro' puzzles here. check it out and tell me how you do https://www.codingame.com/training/easy/the-descent
Wontonimo: you can pick any programming language you like from the dropdown at the top
OofButBetter: i already did that
Wontonimo: and how did you find it? difficult, easy ?
OofButBetter: im on temparature right now
Wontonimo: okay, i made this one https://www.codingame.com/training/easy/fax-machine How hard is it?
Wontonimo: okay, do temperatures
Wontonimo: you get more points for it i think
OofButBetter: ill do it after:)
Wontonimo: do you have a math background or physics? If so I can recommend some that have more of a math or physics style to them
OofButBetter: no not really im fine for now
Wontonimo: :thumbsup:
Wontonimo: Westicles, are you working on anything these days?
Wontonimo: when you say binary search is the big thing, on the twitch stream? It's that ... kinda old school ?
Westicles: I hadn't heard of it before. Seems to be about learning algorithms
Wontonimo: binary search learning algorithms? like a decision tree?
Westicles: oh, I meant https://binarysearch.com/
OofButBetter: well i set up my profile pic
Wontonimo: you are mentally unstable
Wontonimo: (tm)
Thorcode: is there any 13 years old boy from US or UK
Thorcode: pls dm me :)
Thorcode: I need something to prepare for SAT
Thorcode: I need more English word in more subject :)
Westicles: they've dumbed down the SAT so much in recent years that almost everyone aces it
Wontonimo: participation badge
Thorcode: ?
Westicles: back in the old days a 1600 meant you got on TV and invited to space camp
Wontonimo: noone watches TV anymore
Wontonimo: so you wouldn't know even if that still happened
Thorcode: or read news
Westicles: people read news 18 hours a day now
Wontonimo: which people? Fox news doesn't count
Westicles: twitter and whatnot
Thorcode: I can solve SAT if I have more word even the SAT is DUMBED down I can't do it if I don't understand the input :)
Wontonimo: what words are you looking to learn?
Westicles: Just concentrate on writing a nice essay about the important of climate change
Thorcode: the math word
Thorcode: like sum plus minus, etc
Westicles: :thinking:
Wontonimo: https://www.khanacademy.org/ has some great word problems
Wontonimo: that are focused on math
Thorcode: thanks
Thorcode: I can write an essay
Thorcode: Thanks Wontinomo
Wontonimo: let me know if that helps
Wontonimo: you are welcome
Westicles: I always get suspicious when the chat log is missing a day. Something big must have gone down yesterday
Wontonimo: :thinking:
Thorcode: Yesterday my chat broken :(
Thorcode: can't connect to the server :<
C26_1: if stockPrice() > stockPrice(1)
C26_1: I'm just type the wrong button
Wontonimo: good night
Wontonimo: :wave:
Thorcode: Night night Wontonimo :)
TechnoBot: brb
TechnoBot: *brb