Chat:World/2021-08-23
michaellan: stfu
AspectCodin: DRAEM
Retroblaze: brajin kek saathay
AspectCodin: Yenaka nariana vaynu vaynu
Nickname_Already_taken_already_t: i'm batman
hinammmm: i'm batman's dad
cheesecake_assassin: Hi dad
peppers: batman has no dad。
DaBebeWHair: wishing I was batmans dad rn fr
mikhiyoe: me as
mikhiyoe: his bestfriend
AliRazzaghi: im mommy
AliRazzaghi: who want to fuc
AlanSpacer: me
GroutchKeu: na*8 batman
IgnacioGb: Hi, dear community, could anyone tell me which are the fastest ways of getting points in this platform?
Uljahn: what for?
ProCoder03: you will never learn much, if you pursue for just points basically
Uljahn: ye, points here are just means of gamification to slightly motivate you to practice
MrBest: guys please give me sample code of using boost
MrBest: I cant change value of boost...
MrBest: Just can change the thrust
MrBest: cout<<nextCheckpointX<<" "<<nextCheckpointY<<" 100"<<endl;
RoboStac: change the " 100" to " BOOST"
GDEV_09: Anybody knows a way to change the contribute mode (i.e. from classic puzzle to clash) without rewriting the problem?
eulerscheZahl: you can't. delete it and make it new :(
eulerscheZahl: been there myself (puzzle => optim)
Wontonimo: well that unfun
KiwiTae: Kamx drink more water
BlaiseEbuth: Good new! -> That proves you've one!
Kamx: do you get the feeling that you are not good at developping when you are stuck at an easy puzzle ?
Kamx: or its just because im still a student
BlaiseEbuth: I don't stuck on easy puzzles, so no... :smirk:
IndecisionTree: i'd say most knowledge workers in any field get that feeling of "am i actually good enough?" when they get stuck. CS or SWE is no different.
darkhorse64: Close the IDE, take a break, sip coffee.
Kamx: i would understand that , but i think for my case i didnt study/work enough of algorithmics
BlaiseEbuth: No blocking -> No bad feelings -> No life interrogation. Conclusion: Don't get stuck.
Wontonimo: "easy" here doesn't mean "trivial". It instead means that it is easier than the next category and harder than clash of code
KiwiTae: some easy puzzles are not that easy yea..
BlaiseEbuth: Fix your self-confidence: Solve a medium puzzle. This one is easier than most of the Easy-rated ones: https://www.codingame.com/ide/puzzle/rubik%C2%AE
darkhorse64: Part of my language achievements farming list
BlaiseEbuth: Same.
Westicles: Me too.
darkhorse64: with the hard n-queens
BlaiseEbuth: temperature and thor too. Each count for 2.
darkhorse64: A*C, NS, 2048, CG sponsored
BlaiseEbuth: Oo
darkhorse64: ML1
Stella_Artois: does anyone know how to optimize the many horses?
BlaiseEbuth: gg Kamx. Feel better?
Wontonimo: Kamx , this one has a very high community success rate https://www.codingame.com/training/easy/fax-machine
Wontonimo: warning ** self-promotion
BlaiseEbuth: -_-
darkhorse64: Stella_Artois: read the hint
Westicles: And a favorite on fr! "jamais une merde aura ete valide aussi vite ... pffff ... "
BlaiseEbuth: Westicles store compromising data in real time to have mean of pressure anytime against anyone. :D
Westicles: heh heh
darkhorse64: No need for real time. There is an official archive
BlaiseEbuth: I know. But you've to find back what you want. Not sure you can find this specific quote just searching "fax". But perhaps I'm wrong :thhinking:
BlaiseEbuth: (╯°□°)╯︵ ┻━┻
darkhorse64: looking for "merde" is likely more productive
BlaiseEbuth: You'll probably have a lot of results...
darkhorse64: ja_fica: is your C4 bot a NN ?
Wontonimo: why would you want to find swearing by searching for 'fax' BlaiseEbuth?
ja_fica: no
BlaiseEbuth: Wontonimo ?
ja_fica: Its a intelligent opening book
Wontonimo: you said : Not sure you can find this specific quote just searching "fax"
Wontonimo: and the quote had a swear in it
darkhorse64: you are competing with MSmits, then
BlaiseEbuth: Westicle said this quote was about your puzzle.
ja_fica: My AI beats his but its not necessarly better
ja_fica: I dont use meta MCTS
ja_fica: I kind made up my way of doing my own
ja_fica: using some sort of cutted trees
darkhorse64: lots of RPS, you're running me, jacek is doing better and I beat him. So ranking fluctuates
ja_fica: Yes im sorry, im realy ruining the fair ranking for every one :(
darkhorse64: don't be sorry
ja_fica: its because Im finding improvments to my generation algorithm and even testing I'm my computer, I like to really see the results against players
ja_fica: The problem that I dont like about my AI is that is seems that the opening book contains weakness against jacek, mealc, royale and allyourtrees
ja_fica: its unbalancing the ranking...
ja_fica: I'm trying to transition to a NN but I'm having problems in automated data transitions from c++ to python
BlaiseEbuth: "We found how to create a nuclear bomb in our labs, but we wanted to see the results against people"
ja_fica: YES ahahah
BlaiseEbuth: Are you Ok Wontonimo?
darkhorse64: Don't be upset about harsh comments
Wontonimo: thank
Wontonimo: both
ja_fica: RobotStac, does your NN have 10 outputs?
darkhorse64: Yes
darkhorse64: 9+1
ja_fica: 1 for value + 9 for transition probabilities
darkhorse64: That's what he said a few days. He even gave his layers
darkhorse64: I mean the sizes
darkhorse64: *a few days ago*
_Royale: I suppose we can repeat what he said :)
ja_fica: Whats was the size of the input?
_Royale: "2021-08-19 18:03:06 RoboStac nn is Dense layers, 189->128->128->128->9+1"
darkhorse64: 3x63
ja_fica: what do you use royale?
sad_nigglet: idk
ja_fica: 3x63? p0 + p1 + winning position?
_Royale: ja_fica: NN but different layout
darkhorse64: empty, X, O
ja_fica: thats your bot is so strong..
_Royale: ja_fica: I do 126->264->(64+64)->(1+9)
darkhorse64: Both seems equally effective. Cough ...
BlaiseEbuth: Wontonimo: Whatever what you do there'll always be critics, including mean ones. Ignore them, keep the rest. Furthermore, Frenchs love complain about anything. :p
darkhorse64: 264 ? Not 256 ?
_Royale: darkhorse64: I put as much as I could :p
darkhorse64: Power of 2 for computing efficiency
martinpapa69: _Royale you only only dense layers in your c4 bot ? :O
Wontonimo: Thanks BlaiseEbuth. I'm totally overreacting. but there are lessons here for me, like use WIP for a while and consider different programming languages more. Thanks again
ja_fica: it does not make a diference in C4
martinpapa69: use*
BlaiseEbuth: :thumbsup:
ja_fica: power of 2 its just convinient to select but its not optimial
ja_fica: *optimal
ja_fica: I do have implemented an AVX trick that grouped with a NN would be so worth to see
oddly: how do you finish the first ai quest?
_Royale: martinpapa69: yes only Dense with relu, and tanh/softmax for the outputs
martinpapa69: wow, thats surprising. I thought it could only work well with few conv layers
ja_fica: i think sigmoid for the 9 layers also work
BlaiseEbuth: oddly what's the first quest ?
ja_fica: well conv are expensive in time, good for memory... shared weights can be a good mid term
_Royale: martinpapa69: I don't know how to code do the conv layers yet
ja_fica: share the weigths in patchs of neighbor pixels
martinpapa69: I already have a convnet for c4, but it doesnt work at all. rank40...
_Royale: I also tried to put the holes as inputs (189 like robo) but it did not seem to improve for me
ja_fica: I would'nt do it either
ja_fica: im think using only 63+8*7 as input
_Royale: ja_fica: and I don't do AVX, but I tried to have good autovectorization
oddly: so ya the first ai quest is a mission for me, i have to fix the cords and adjust the speed
ja_fica: without my AVX my roullouts fo from 65 to 31
ja_fica: *go from
ja_fica: 65k to 31k
ja_fica: 63+8*7 is enough for the neural network, current player + one-hot-encoded next possible play
ja_fica: 119 inputs that can contain more area related info for the transition probabilities
TobiasA: does "int* x" do the same thing as "int *x"?
martinpapa69: :thumbsup:
TobiasA: thanks
BlaiseEbuth: int *x-> Ugly int * x-> Weird int* x-> Perfect Though they all do the same. ^^
sad_nigglet: ok
TobiasA: nothing worst done "synonyms" in programming
TobiasA: *than
BlaiseEbuth: Not really synonyms. It's more conventions. Like: http://chat.codingame.com/pastebin/a1c2a33a-859a-4e06-a5cf-b32dda0fa619
BlaiseEbuth: You can write "int*x", that's correct. Just a space story
martinpapa69: I've heard that in some software companies you get instantly get thrown out of the window for that
BlaiseEbuth: I totally approve.
TobiasA: yea that looks disgusting
Westicles: defintestrate?
TobiasA: google doesn't seem to know what that word is
TobiasA: neither do i:sweat_smile:
Westicles: just a pun on defenestrate
TobiasA: oh eye see
TobiasA: you see what eye did there:joy:
martinpapa69: icy
TobiasA: ice sea
jacek: oO
BlaiseEbuth: (╯°□°)╯︵ Oo
oxydowe: icy
oxydowe: martinpapa69 i didnt cy your previous message
TobiasA: you mean eye didn't see:joy:
oxydowe: yes, Heidy dnt sea
ArmanTaheri: hi
ArmanTaheri: i'm new
BlaiseEbuth: Hello new
ArmanTaheri: can we be friends i think
BlaiseEbuth: Moderator!
ArmanTaheri: oh kind of rude
jacek: Diverator!
BlaiseEbuth: Astrobytes!
BlaiseEbuth: Help!
ArmanTaheri: be friendly lads
TobiasA: spawn_moderator()
TobiasA: that's how to do it
Astrobytes: he's gone. Chill.
ArmanTaheri: can any one know how to do some basic challenges ?
ArmanTaheri: i just see some advance on
BlaiseEbuth: :relieved:
ArmanTaheri: ok i think you dont care
ArmanTaheri: have a good time
AllYourTrees: :scream:
oxydowe: has anyone experienced strange behaviour with notification? like: click to close one and all of them are gone?
charqwer: yes
Mahin1729: I clicked to close one, and all of my chrome tabs were gone
charqwer: nice
oxydowe: bug report? how?
Scarfield: on the discord server, there is a bug report channel
oxydowe: ty
jacek: discofield
Scarfield: ElysiumCek
RoboStac: (I know this is responding to an old message) - the empty squares are in my input because otherwise the conv layers struggle at the edges of the board and I never changed it when I dropped all the conv layers
RoboStac: they probably don't do much at all on the current version
jacek: using onehots in c4?
martinpapa69: cout << ( 2 + 3 & 1);
martinpapa69: 1
martinpapa69: :scream::scream:
jacek: is it wrong?
martinpapa69: my whole life's been a lie. i thought bit operations have the strongest precedence
AlanSpacer: And I thought my parents loved me
JimmyJams: and I thought my gf was a woman
ja_fica: jesus
ja_fica: I still suck implementing NN in c++, but I wont give up until I do, then I guess it will be copy paste to UTTT...
jacek: NN in UTTT? :scream:
ja_fica: why not?
jacek: sure, the more the merrier
AllYourTrees: do ya'll hvave your NNs output value in [-1, 1] or [0, 1]?
jacek: [-1, 1]
jacek: what abomination is [0,1]
AllYourTrees: :shrug:
RoboStac: [0,1], who wants to deal with negativity
martinpapa69: -π , +έ
TobiasA: style points
AllYourTrees: anyone compared [-1, 1] to [0, 1]? does it impact anything?
RoboStac: it doesn't really matter other than adjusting exploration rates, but I use negatives as solved values
jacek: youll have to change your hyper parameters all over again eh
Wontonimo: i use [0,1] for uttt and c4
kakben: anyone used monte carlo tree search to solve mars lander?
kakben: if so, what did you use for your scoring of the end states from rollouts?
Wontonimo: which mars lander episode?
kakben: episode 2. i realized now something might be wrong with my physics sim. how does the game handle integer rounding each round?
darkhorse64: Internally, the game keeps floating point values
kakben: ok so the display in integers is a simplification of reality then
kakben: thanks
darkhorse64: Yes
Stilgart: would be nice to have access to the referee for mars lander
borrel: i have heard that the first round vectors are all rounded
kakben: no rounding, i solved it now. my problem was not adding half of the acceleration to the position update
Astrobytes: that helps ;)
ja_fica: I tried [0,1] in UTTT
ja_fica: I had to lower the exloration adjust the constant
ja_fica: *lower the exploration process
ja_fica: *constant
kakben: mine is fine at around 2, but i have trouble getting my lander to select the correct angle for its final move. no idea why yet
Wontonimo: the final angle and speed have to be part of the scoring
Wontonimo: alternatively you could instead target just above the landing pad with your MCTS solver, and then switch to an if/else bot for the final 50 units of height
ja_fica: I'm proud of my last submition..
ja_fica: in C4, bet almost 2th
AntiCheap: hay any tips for the resistance puzzle? I'm using a tree but it's not efficient enough both js and rust.
AntiCheap: I also tried removing some words of the dictionary filtering ones present in the morse code input.
ja_fica: its probably a complexity issue
ja_fica: make sure you are not revisiting states
viper5466: this stuff is too hard for a beginner :o
Wontonimo: yes, this isn't a beginner site
Husoski: Does anyone know what makes "View Contribution" appear in the classic puzzles? I've only noticed it recently, so either I crossed some magic line or there's been some change to the software. (..or both, I guess.)
Wontonimo: i have no such thing. where do you see this ?
Husoski: Open up a puzzle to see the description, click Solve to get to the IDE for that puzzle and I see it on the right side of the banner at the top of the problem statement (yellow, all caps)
Husoski: Wontonimo, maybe it's a level 31 thing? I just got there about a week ago...
Husoski: Hmm. Not seeing it on the game-loop puzzles like Mars Lander, though.
ProCoder03: probably, Its only for community puzzles !?!
sad_nigglet: mm true
JimmyJams: you ever just write 10 lines of code and it just works the first time somehow and you're like "wtf there must be something wrong here" and you spend 20 minutes looking for a flaw that isn't therE????????
martinpapa69: no
AlanSpacer: cant say that i have
JimmyJams: me neither
JimmyJams: I feel bad for the losers that do that