ItsNotABug: In the entry code review, can I select neither? :'D
antiwonto: [auto] Hey Tien118, here is a :taco: for loggin in today while it is quiet. You now have 1 tacos
antiwonto: [auto] ':taco:' was defined as ' A currency to some, sustenance for others '
BenjaminUrquhart: "back in my day, the only chat bot was Automaton2000"
Automaton2000: im not even sure if it is in the top league
Urutar: Automaton2000, you certainly are :D
Automaton2000: you probably have to rewrite my bot
Urutar: Which instructions?
Urutar: Entry code?
Urutar: Ehhh, there are far worse ones
Urutar: At some point, you'll be able to^^
Urutar: Nah, sometimes they just have hard to grasp mathematical or technical ideas behind them
Chris_C769: agree on the math one sometimes that true but other times there are terrible instructions I'm doing one where you have to shoot an enemy no instructions only one way to do it yet I'm wrong. Makes no sense. It tells you select closest enemy by distance. Yet clearly that is not what you are to do
Uljahn: "There's no accomplishment without struggle. No triumph without tests and failures along the way."
LelouchVC2: I accidently discovered the cure for cancer
antiwonto: [auto] Hey LelouchVC2, here is a :taco: for loggin in today while it is quiet. You now have 1 tacos
LelouchVC2: ;D !
antiwonto: [auto] Hey gokubill, here is a :taco: for loggin in today while it is quiet. You now have 1 tacos
gokubill: cai j vay
gokubill: me may
dot.nick: just hit rank #69 in clash of code. nice
derjack: That. - she
derjack: happy friday the 13th
yujin0806: wish us luck:joy:
MajorKurk: If you see someone with an axe, don't run, it's too late.
MajorKurk: Friday the 13th x) It's safety recommendation.
MarinDeNsi: according to you, which programmation's language is the one with the least size of syntaxe ? thx by advance for answering !
BobLob: There are several Turing complete languages with 8 instructions ( I made a variant with only 1 instruction, but it was cheating by abusing MOV ). They're usually considered toy languages, though
BobLob: For a language actually in use, C has 32 keywords. I'm not sure if it's the smallest by that metric, but it has to be close
Felikan: Freedom for Scotland
- plays bagpipe aggressively*
Astrobytes: dbdr, Tiramon, mods to #de and #world
Felikan: I mispelled "How to center a div" sorry pls dont ban :(
Felikan: My cat walked over keyboard
nicola: Sorry, mine slipped over mine too.
Manchi_o6o7: if someone could help me out with a cpp question
Manchi_o6o7: https://pastebin.com/9tEk0D0Z Machine is the abstract/base class, but I want to have this method working for all the child classes. The problem is, when I define it as an abstract class, I can't construct the Machine object in my function parameter list.
Manchi_o6o7: Is there any walk around
wlesavo: you cant construct objects of the abstract class, thats the whole point, but you can make an object of a child and create base class pointer from it
Manchi_o6o7: But is there any way to achieve this funcionality. where I want the Machine object basically to be some_derived_class
Manchi_o6o7: so I don't need to define multiple times methods that are the same, for every derived class
DaNinja: should work if makeIntersection is public
Manchi_o6o7: maybe here it is better explained
jzen: I think you need to pass other as reference in the capture clause?
jzen: [this, other] -> [this, &other]
jzen: I am not sure if I understand
antiwonto: [auto] 'help' was defined as ' https://how-to-properly-ask-a-question.surge.sh '
nicola: SATANSPAWN, I’m the cacodemon.
5DN1L: English only please
5DN1L: and mind your language
Astrobytes: oh nice, afk for 2 hours and now I'm pushed to legend in spider attack :tada:
darkhorse64: grats: Astrobytes and your rank is quite high. Any tips ?
Astrobytes: thanks darkhorse64! Hm, really nothing special, didn't use control at all. Defender and farmer do the same thing except defender stays a little closer to base, only use wind at a certain distance from base. Attacker immediately to enemy base and wind as much as possible
Astrobytes: Also attacker uses shield
Astrobytes: It's the same code from the contest, just with a few values changed here and there :D
Astrobytes: (all my rewrites failed)
Astrobytes: I have to say it's not at all a good bot, yet effective enough.
5DN1L: You too, Aghiles09
5DN1L: np :)
5DN1L: why would someone go to a place and randomly ask others to shut up? how weird
darkhorse64: I have found that using spawn points as guardpositions for my farmers is good
darkhorse64: at least for the initial phase
Astrobytes: Yeah I had a version doing that last night, was quite effective but it broke my submit higher up in the league so I reverted
Astrobytes: Now I don't know if I wish to keep working on it or move on to something else
darkhorse64: maybe try to implement wind cannon just for fun
Astrobytes: I think I might start by refactoring a little, it's only 500 lines but it's horrible
darkhorse64: Sort out before expansion :relaxed:
mounir.: hey bitchis
Uljahn: both kicked
Wontonimo: thanks, i was just about to
Astrobytes: thanks :)
Darleanow: HELLO WONTO
Darleanow: HELLO ULJAHN
NewCoder09: All of my state test are finally done
NewCoder09: i have exams next week
Wontonimo: hi Darleanow and NewCoder09 :wave:
Ayza: is stealing the boss's strategy a good idea lol
Ayza: sparing one hero at the enemy's base to use control on the enemy heroes
Astrobytes: control isn't even needed, just winding spiders at the enemy base is enough
Ayza: you really got to legend without it? damn
Ayza: gotta try that
Astrobytes: It will depend on your defence/farming strategy too though, everyones code is different ;)
Astrobytes: ideally I'd have been using control + wind but I'll do that after the rewrite perhaps
PrzemekS: guys i got math problem. I made dot product of two normalized vectors (monster movement vector and from monster pos to my base vector) . When dot product reaches 1 i know monster is going straight to my base, but what about the minimum (to reach base radius), i don't want to guess 0.6 or something , any math way to do that?
Uljahn: you mean ray-circle intersection?
Wontonimo: I do it the dumb*ss way, just simulate a few dozen moves and test
Wontonimo: also, if the mob is within your base radius it is always aiming for your base
Wontonimo: lmgtfy ray-circle intersection
antiwonto: [auto] try 'lmgtfy [username] search'
Uljahn: lmgtfy [username] search
antiwonto: [auto] try 'lmgtfy [username] search'
Wontonimo: lmgtfy lmgtfy ray-circle intersection ray-circle intersection
antiwonto: [auto] try 'lmgtfy [username] search'
Wontonimo: lmgtfy ray-circle intersection ray-circle intersection
antiwonto: [auto] try 'lmgtfy [username] search'
Wontonimo: (╯°□°）╯︵ ┻━┻
Wontonimo: sorry for the spam
Astrobytes: lmgtfy PrzemekS ray-circle intersection
antiwonto: [auto] hey PrzemekS let me google that for you https://letmegooglethat.com/?q=ray-circle+intersection
Uljahn: the bot is too smart
Wontonimo: i obviously can't use ctrl-v and my own bot
5DN1L: that's why it's called antiwonto, Wontonimo :wink:
Darleanow: waddup wonto
Darleanow: Helo 5DN1L, glad to see u more often
5DN1L: Hello Darleanow :)
eulerscheZahl: if you already like 'u', you will love 'w'
MSmits: so I did a lot for my PR-2 bot. I did symmetry detection, BFS on all hexes, I put in hungarian method, the contamination method to deal with enemy and fog of war. But my bot still doesn't move :P
eulerscheZahl: submit anyways?
MSmits: well I won't crash at least :P It just does wait
Wontonimo: PR-2 ?
eulerscheZahl: platinum rift 2
Astrobytes: lol, definitely a bug in the game :P
eulerscheZahl: or pull request
MSmits: I vow I will have a working bot this weekend :)
MSmits: I found this btw euler: https://github.com/antifriz/hungarian-algorithm-n3
MSmits: is that the one you used?
MSmits: seems to work fine
eulerscheZahl: i found 3 implementations. first was too slow. 2nd was fast but gave wrong results. 3rd was in java so i had to port it to C#
eulerscheZahl: i linked it in my code of ice and fire post mortem i think
MSmits: ohh ok
MSmits: will check it out
5DN1L: kristos44 yes, if you play the code once then it's saved
eulerscheZahl: if you indent with only 2 spaces, i hate you
MSmits: I tab
Wontonimo: i feel your hate. it won't change me
Westicles: two spaces and no blank lines is the obvious way to go.
5DN1L: np :)
antiwonto: [auto] ':upside_down:' was defined as ' ɥǝllo '
ethaneCH3CH3: take taco
antiwonto: [auto] That taco command has moved to the #taco channel
Astrobytes: MSmits: from that java code... "matchJobByWorker"
Astrobytes: see anything amusing?
Westicles: look at this beautiful solution
MSmits: lemme see
Astrobytes: you don't have to look, just that name
jacek: whats wrong with it
MSmits: ohh ok
MSmits: it's not that strange? I have longer names sometimes
Astrobytes: The middle two words
MSmits: ohhh! Poop!
Astrobytes: Jobby worker
MSmits: Astrobytes finds the poop in any program
Astrobytes: jacek: jobby in Scottish means a :poop:
jacek: those scotts eh
jacek: well all i can say is lul
Astrobytes: Like one of yours MSmits
Astrobytes: except in Java
eulerscheZahl: i know the website where you got that from jacek
jacek: no wonder, i linked it
Astrobytes: the generator one?
Astrobytes: ah so you did
eulerscheZahl: oops, reading chat partially was never a good idea
eulerscheZahl: i saw taco and skipped the next few lines
nicola: He insulted me in PM.
Wontonimo: good to know. thanks
Westicles: I've been cleaning up and publishing old puzzle solutions. Weird that all the first ones I did use random numbers to solve
Westicles: I think SA was the only approach I new so I used it even on easy puzzles
Darleanow: ok guys i'm retarded, i was talking to the Clash of code bot
Darleanow: because i got 50% and him too
5DN1L: Next time talk to Automaton2000 instead, at least it will respond to you :)
Automaton2000: do you know what i need to change the value of log(parentvisits), i recompute it every time, perhaps i should ...
Darleanow: Hello Automaton200 :)
Automaton2000: if you have no idea why
Darleanow: H0o0w0 a0r0e0 y0o0u0 ? Automaton2000
Automaton2000: what would you do that
Darleanow: How are you Automaton2000
Automaton2000: this is the right one
Darleanow: ANSWER ME
5DN1L: it knows the second "how are you" is the right one, not "H0o0w0 a0r0e0 y0o0u0" :wink:
Darleanow: Okay i see :)
jacek: no, its community game
Wontonimo: I've never seen that game !?! Very cool
Wontonimo: is the source code available for Code Keeper - The Hero ?
5DN1L: look for the link in your browser history
5DN1L: Is the clash finished?
Manchi_o6o7: Hey guys. I wrote some C++ code which builds me a custom URL to represent a graph I am working with, on an online visualization tool. Now I am wondering, is there any way to invoke the download of a svg. from the page? I will paste the link of the site, so you can see what it has to do with.
Manchi_o6o7: This is my code, but it isn't working because it isn't accessing the direct download link.
Uljahn: search it here http://eulerschezahl.herokuapp.com/codingame/puzzles/
Uljahn: there's no official way to go back
NewCoder09: Hi guys
Manchi_o6o7: Is maybe someone of you familiar with web programming
Wontonimo: yeah, i've done programming of webs. what kind of webs?
**Wontonimo you are in a dungeon corridor. Mildew fills the air and rickety beams hold the earthen walls from collapsing. Ahead you hear laboured groaning of something large. Your torch light is sputtering.
Wontonimo: P0sandu roll for initiative
P0sandu: gm == good morning
jacek: Wontonimo code keepers looks more like optim than multiplayer game
Wontonimo: i noticed right away, but it prevents people from hacking the hidden test cases I think.
struct: Are the replays laggy for anyone else on that game?
DomiKo: you mean code keeper? then for me nope
jacek: trying dice duel?
DomiKo: dice duel is smooth too
jacek: lol, it crashed
jacek: java.lang.RuntimeException: Error reading referee command. Buffer capacity: 30000 / 30000
jacek: darn i replayed and didnt save replay
struct: dice duel is fine for me
struct: but not code keeper
struct: which game jacek?
DomiKo: Code Keeper 100%
jacek: referee once crashed
struct: I think that error is when you send too much info on inputs
struct: you have a limit of chars
DomiKo: I guess its rendering problem
DomiKo: because there is a lot of animations to do. and aCat divides turn into many turns just to do animations
jacek: this https://www.codingame.com/replay/634733366
jacek: its weird
DomiKo: I can't see it :(
eulerscheZahl: btw i updated the code keeper map generator. i don't want to ping aCat a 3rd time about it
ImSeaFu: hiya jacek
ImSeaFu: I've been told you're someone I should get to know since you're good with ML xD
DomiKo: euler should I tell him?
ImSeaFu: So hello!
eulerscheZahl: yes :)
jacek: my Mars Lander is just bunch of ifs ;d
eulerscheZahl: he has the link already. i also shared it on the forum
DomiKo: I will tell him tomorrow then :D
eulerscheZahl: real life pings can be more efficient sometimes
DomiKo: yeah it will. He told me sometime days ago, that he should look at yours generator
DomiKo: some days*
eulerscheZahl: maybe i'll find some motivation to give the game a try then
eulerscheZahl: it looks really cool
DomiKo: it's really cool
jacek: ImSeaFu any question?
eulerscheZahl: but somehow, between real contests, i didn't get to play it :(
ImSeaFu: rofl wow I'm dumb I thought you meant that your Mars Lander is shitty machine learning but I get now that you were subbing for ML xD
ImSeaFu: I mean about a bajillion but I gotta do some more studying. Just making myself known so that when I get hard stuck and need help that you at least have heard from me before xD
ImSeaFu: I've only really done the Google ML Crashcourse and am now in your Neural Network Xor example from scratch xD
ImSeaFu: So I got a ways to go xD
ImSeaFu: but machine learning is really what I'm most interested in learning so gotta chip away at it xD
jacek: are you scarfield by any chance
ImSeaFu: But after doing the google crashcourse I was kinda just sitting there going like "great so now I know a bunch of theory, and have some pre-written code, but I don't feel like I even know where to begin with actually using it"
ImSeaFu: maybe, I don't know what a scarfield is so it's hard to say xD
jacek: ok so youre not ;d
ImSeaFu: Well that's a relief
jacek: do you have some programming background?
ImSeaFu: I've been programming on and off extremely poorly for 20 years
ImSeaFu: so not really
DomiKo: 20 years damn
ImSeaFu: Just over the last few months have I really started taking it more seriously. My fundamentals were/are really bad because I have/had a bad habit of just rushing to higher end things without truly understanding it
jacek: so machine learning in essential
ImSeaFu: Picked up Python since I heard it's one of the leading languages for machine learning
ImSeaFu: No, learning everything, lol I'm a really bad programmer, 20 years or not I did not study well, practice much, or have good habits at all
ImSeaFu: you should pretty much consider me a newbie. Maybe like 6 months of actual knowledge
ImSeaFu: Like I still struggle with classes and basic shit xD
antiwonto: [auto] ImSeaFu be respectful and watch your language https://www.codingame.com/playgrounds/40701/help-center/code-of-conduct
ImSeaFu: bad language
ImSeaFu: But yeah if anyone has any advice for things I should look at to help my study that'd be cool, for now though I am going through this playground and seeing if I can understand it xD
ImSeaFu: hmmmm what is for _ in range(HIDDEN): for example? like what is the _?
jacek: normally youd write for i in range(), but since i is not needed, its replaced by _
Disa86: @ImSeaFu "_" coube "i" or "a", it's just a name for the item you are currently operating on in the iterable "HIDDEN"
ImSeaFu: so it's basically just like a token that says "idgaf what this is?"
DomiKo: _ in python is like a trash can
ImSeaFu: ahhh kk
jacek: so this for loop will run for HIDDEN times
ImSeaFu: that's what I was confused about, I've seen i and other variables but _ isn't a valid variable name afaik so I was confused
5DN1L: you can change _ to anything you want
jacek: ok, ill change it to range
5DN1L: not reserved words...
ImSeaFu: I'm going to change it to HIDDEN
5DN1L: for HIDDEN in range(HIDDEN)?
5DN1L: no, that won't work
Disa86: @ImSeaFu If you're looking to learn the basics of machine learning and you wanna start with Python I highly recommend "Neural Networks From Scratch in Python" by Harrison Kinsley and Daniel Kukiela
ImSeaFu: it'll create a blackhole and explode I'm sure of it
5DN1L: yup :smirk:
Disa86: it's a great book for beginners
ImSeaFu: Book is really the worst way for me to learn
ImSeaFu: I have very bad ADHD so keeping myself focused on a book is tough
Disa86: it's complete with exercises for you to follow along, and QR codes to youtube animations to help you understand
ImSeaFu: now you're speaking my language
Disa86: honnestly one of the least boring programming books I've ever read lol
ImSeaFu: Hmmm so maybe this is a dumb question but I'm not what you'd call a big "reader". if I get the eBook can I read that on my PC easily or is it gonna be a big pain in the ass?
Disa86: There's a PDF version only on the official website if you want
struct: I bought the pdf
Disa86: nnfs.io is the site if I recall correctly
Disa86: yup it is
jacek: struct you bought from smits?
struct: no, from the site
Astrobytes: lol jacek
ImSeaFu: hmmm I can only find the paper version and the ebook on nnfs.io
struct: ebook = mailed pdf
struct: I think
Disa86: yeah it's a PDF
ImSeaFu: ohhh okay
struct: ImSeaFu you can also check his videos first
ImSeaFu: his who? 2 people wrote this book from the looks of it xD
eulerscheZahl: those linking skills
ImSeaFu: wait a minute.... it's sentdex?
ImSeaFu: Lol I know that guy, I'm in his discord xD
antiwonto: [auto] Hey Astrobytes, here is a :taco: for loggin in today while it is quiet. You now have 43 tacos
Astrobytes: :taco: jacek
antiwonto: [auto] Astrobytes has awarded jacek 10 tacos. jacek now has 32 taco. Astrobytes now has 44 taco
PatrickMcGinnisII: today is me and my GF like 4 yr anniversary. i asked her if I should make the bed (we don't live together), she said no. Should I assume that she's no gonna put out? And should I dump the B? tmi, i know.
**PatrickMcGinnisII is suffering from a lack of tacos
Astrobytes: Just enjoy the anniversary and have some fun. You know your GF better than us (well, hopefully) so you judge the rest lol
PatrickMcGinnisII: doing save the forest puzzle again, why do i want to let it burn? ;)
Astrobytes: Sounds like you need a new GF
PatrickMcGinnisII: College friend is dying in a hospital in Lauderdale right now, I wanna say fk the anniversary and go see if I can say goodbye...maybe that's what has me messed up
Astrobytes: Dude, go see that friend. You only get one shot
Astrobytes: If she's a good one, she'll understand
PatrickMcGinnisII: logic, who knew u could find it on CG
Astrobytes: Go donate $5 to a good cause. That's my fee :rofl:
jacek: good cause, like bioengineering
PatrickMcGinnisII: this is where i say something messed up like: $5 only gets you a virtual hooker. smh.
PatrickMcGinnisII: talk about bio engineering
PatrickMcGinnisII: Ok, a woman just lost her baby near here, I'll drop it in her gofundme
PatrickMcGinnisII: there's a gremlin in the IDE
Astrobytes: good lad Patrick
PatrickMcGinnisII: just turned this
PatrickMcGinnisII: I'm done for today, glhf
Astrobytes: I thought it was some kind of opening book lol
PatrickMcGinnisII: it could be
Astrobytes: hope you see your friend dude, take care
jacek: opening books are more like this 搯擋擾撻攕攥断旝旋昁晕曐旴机暩晞曩曡杋憇戲斔扲斔揞摸擆擐擉摻搏揆搅捤掶撐撍斉斀暕晗曢柪枡帢椙暉泼柫昚斁攉揚
Astrobytes: of course, but you know what I mean
Wontonimo: oh, that opening book. i know it well ;)
struct: my openings book usually fit in 64 bits
jacek: the story, characters development and that plot twist!
Astrobytes: Enough chars for even M. Night. Shyamalamalamanalamanna... I give up
Astrobytes: (re. plot twist)
Astrobytes: Incidentally: M. Night clearly stands for Mad (k)Night. Silent 'k'.
jacek: the inventor of mad pod racing?
Astrobytes: "inventor" :D
Astrobytes: I think it was Squid who invented that name iirc
Astrobytes: then Tibo ran with it when it came to changing the names
jacek: :taco: derjack
antiwonto: [auto] jacek has awarded derjack 10 tacos. derjack now has 50 taco. jacek now has 33 taco
Wontonimo: more farming eh jacek ?
Wontonimo: I never thought anyone would be farming tacos in my life
Astrobytes: via *your* taco bot
jacek: need more balls https://streamable.com/qgza5b
Astrobytes: BR, nice. Did you succeed in SR?
jacek: darn openai evolution strategies are lit. my crappy GA cant compare with them. training in minutes instead of hours
jacek: not yet, still figuring out how to evolve and coevolve with opponents
Astrobytes: ahh, cool project though
jacek: im hoping to create successive training framework for such games like i already have for board games. so more leaderboard to own :imp:
Astrobytes: More power to you if you can succeed man, I think it's a nice evolution of the platform for lots of NN/ML action
struct: why 2 balls on first frames?
jacek: there are 2 balls in br2k
jacek: the nn inputs see only one ball at one moment
struct: you are right
jacek: unless the car carries one
struct: I just didnt see it in the other frames
struct: due to the transparency
jacek: this is nn. two ouputs - thrust and angle. even if i have decent blackbox that tells me the action, i dont know how to use it in search yet
Astrobytes: nice guess
jacek: i judge from code :v
Astrobytes: kinda obvious now I read it :D
struct: why is it capslocked after the _
Astrobytes: easy enough to fix
Astrobytes: Or keep it, but -1000 style points
Wontonimo: you can use an attention network to identify the ball to focus on jacek
jacek: well the problem with br2k how to do inputs right. for example 2 balls, inputs should be order invariant
Wontonimo: heck, the attention network could even order the balls so that the balls are order variant based on importance
Wontonimo: and it will learn to place the balls in the correct input
jacek: any tutorial i could read on this?
jacek: keep in mind right now im training via neuroevolution, randomly perturbing wights instead of some smart q-learning thing
Wontonimo: hmm... the attention network is just an eval per ball. given the balls data and the player, is the ball important?
Wontonimo: very much like your other NNs, but this NN is applied to each ball, and you sort the balls based on that eval
jacek: NN for sorting the inputs?
Wontonimo: imagine you have data for each ball, let's call it matrix [X]
Wontonimo: score the ball given M2*tanh(M1*X + B1) +B2
Wontonimo: do this for all balls
Wontonimo: sort the balls given that scoring
Astrobytes: there are only 2 balls
jacek: it could be 1 or 0 if someone has it
Wontonimo: oh, that makes it easier
Astrobytes: BitRunner 2048
Astrobytes: true jacek but the balls technically still exist
Wontonimo: X is 2d, so let it have 2 rows
Wontonimo: and the output is 2 values
Wontonimo: 1 for the 1st ball, 1 for the 2nd ball
Wontonimo: Dense(Dense(balls data))
Wontonimo: where the final output is 1 value per ball
Wontonimo: the ball with the higher output is the primary
jacek: and based on this put it into main network?
Wontonimo: yeah, have the higher ranked one go to the 1st position in the main network.
Wontonimo: once you have that working, make room in your main network to accept a second ball and put the second ball input into it.
Wontonimo: treat the second ball data as if it is something completely different and not a ball. just 'additional data'
jacek: ok. if this works, i can do that for order of enemy cars as well.
jacek: thanks. ill have somehting to work on
Wontonimo: cool! let me know how it goes
Astrobytes: that's an interesting idea, I'm no NN person (yet) but I like that
Astrobytes: and what of the situation where an enemy car is carrying a ball? Or any car for that matter
Wontonimo: I use that for a chatbot I developed at work for it to decided between different response handlers. It's the bomb
Astrobytes: nice, lightweight too
jacek: Automaton2000 can i have your attention?
Automaton2000: i just don't know it
Wontonimo: for the chatbot, a particular response handler has some phrases that it is looking for like "how far is the moon" and "help" and "what is the correct mortgage calculation for including home equity loans". It uses cosine distance over known matching phrases to see if it is the right handler. Then the right handler will try and solve the problem.
Wontonimo: This lets me add more handlers without having to do anything with the previous ones or re-balance some decision tree
Automaton2000: i made a bot for a game like that
jacek: oh my
**TKirishima is impressed
Uljahn: cosine distance means word2vec but for phrases?
Astrobytes: that was my understanding
ImSeaFu: @wotonimo 5DN1L recommended your YT channel for ML, any recommendation with which video to start with?
Wontonimo: yes Uljahn
Wontonimo: my channel for learning ML? No, there are only a few videos nothing like a course.
jacek: you can start with https://www.youtube.com/watch?v=dQw4w9WgXcQ
Wontonimo: oh, sorry, i re-read you question Uljahn and not that' it's the right thing
ImSeaFu: jacek -.-
ImSeaFu: I know that URL anywhere
Wontonimo: oh man, i fell for it
Wontonimo: ImSeaFu , check out this set of videos ! https://ocw.mit.edu/courses/6-034-artificial-intelligence-fall-2010/resources/lecture-1-introduction-and-scope/
ImSeaFu: I feel like in trying to learn this whole ML thing I just end up going over the exact same info over and over again xD
jacek: put it into practical use
ImSeaFu: I'm so far from being able to do that
Wontonimo: what do you want to do with ML ?
ImSeaFu: I mean lots of stupid stuff, but obviously you gotta learn the basics before that xD
Wontonimo: for example, let's say you have written some tricky function that returns a nice Eval for Spring Challenge 2021
jacek: i was doing much stuff before i understood ml
jacek: frankly, the more i work with NNs the less i understand how they work
Wontonimo: and this eval you have written has a lot of magic numbers in it, like constants you've hand tuned
ImSeaFu: One of the things the piqued my interest a few years back was SethBling's MarI/O program that learned to play 1-1 on it's own
jacek: how about this mario with GA https://www.youtube.com/watch?v=CI3FRsSAa_U
Wontonimo: well, you could make all those magic numbers be variables (we call then hyperparameters because we are fancy) and you could use something like evolution to discover the best hyperparameters
jacek: it even learned to wall jump
Wontonimo: that's effectively what jacek is doing with his neuro evolution
Wontonimo: in his case the entire NN is treated as hyperparameters
ImSeaFu: I mean I feel like I'm not even close to being able to do that. Like it feels like there is a step missing in between where I am and where I need to go but I have no idea what that step is
Wontonimo: okay, here is a super super intro https://www.youtube.com/watch?v=wUPFOFbCeFI . Tell me what you think
Astrobytes: (you pushed him hard enough to recommend his own video! More of that!)
Astrobytes: It's a great video.
Wontonimo: :taco: Astrobytes
antiwonto: [auto] Wontonimo has awarded Astrobytes 10 tacos. Astrobytes now has 54 taco. Wontonimo now has 13 taco
ImSeaFu: Cause I went through the Google ML Crashcourse and went over a lot of the terminology, techniques for getting lower loss, etc. But all of the code was provided to me, I was essentially just messing with some dials which needless to say is much different than being able to build it on your own xD
ImSeaFu: Well I already feel discriminated on by your video
Astrobytes: #taco channel should be called #TacoHell
ImSeaFu: "Have a clear mental image of how ML works"
Wontonimo: What? how did my video discriminate? I really watch out for that stuff
ImSeaFu: All well and good for people who don't have aphantasia like me :P
Wontonimo: oh. i have aphantasia also
Wontonimo: i don't mean mines eye mental image
ImSeaFu: Lol I was just messing with you, I mean I do have aphantasia but I'm not actually offended :P
Wontonimo: i always thought was was poetic until just a few years ago i discovered people actually could 'see' things in their mind
jacek: apparently lucid dreaming isnt affected by aphantasia. people with aphantasia report on LD forums they do fact have lucid dreams and theyre vivid
Wontonimo: weird to think people are walking around hallucinating all the time
ImSeaFu: lol yeah.... no one ever told me that normies can actually see things when they close their eyes xD
Wontonimo: yes, i can control my dreams. but also sometimes my dreams are super abstract wiithout dimensions
jacek: the imagea are fogg, fragmented mostly
Wontonimo: My wife and kids are on the opposite side of the spectrum and can for example count the spots on the family cat from memory
Astrobytes: I don't have full-on visions when I try to visualise things, very far from it, nothing 'visual' at all mostly, but it's *there*
Astrobytes: and I have a strong imagination/creative side
Astrobytes: I believe I am in the middle
jacek: replay experience buffer is like imagination isnt it
Wontonimo: So there are at least 2 separate areas in the brain for remembering visualish stuff. one that is 2d-ish and colour, and another that is 3d-ish and conceptual. The second is usually considered spacial awareness, it allows you to know there is a chair behind you without having to 'see it in your mind'
Wontonimo: i have a stong 3d spacial awareness but no iiimage/colour thing
Astrobytes: I'm colourblind too so that probably doesn't help
Astrobytes: also: "colour" :D :tada:
jacek: no voices either?
Wontonimo: no, only my own
Wontonimo: even when remembering a song, it is my voice, even for the instruments
Astrobytes: That's interesting, for that kind of thing I can definitely get other voices
Astrobytes: is there a left-brain/right-brain thing here?
Astrobytes: or hemisphere if you want to be correct
Wontonimo: i don't think so. i think it is a what-you-practice-when-young thing
Astrobytes: I've been cursed with creativity and logical leanings
Astrobytes: so I've always been into art, music and technical stuff
Astrobytes: nightmare at school, nightmare in life
Wontonimo: when i was practicing mnemonics a lot it seemed like I was getting more detail in my minds eye and not just concepts being placed in space but just wisps of detail like textures and noise. it would surprise me
Wontonimo: when i stop practicing every day it went away
Wontonimo: yeah, me too Astrobytes, but not having a minds ear music has been difficult
Wontonimo: Hey ImSeaFu , did you find the video helpful?
Astrobytes: so many methods and so many wrong methods of retaining/absorbing information and *there is no one correct way*
Astrobytes: Since we don't fully understand how it's done, and indeed if it is done in the same way in different individuals (although similarity in region of brain etc is obviously proven)
jacek: maybe you just have too little learning rate
Astrobytes: it is hard to discern some foolproof method
Wontonimo: i've often that was my issue jacek ;)
ImSeaFu: Wontonimo I just finished watching it and I'd have found it helpful if this wasn't the same concepts that I've gone over in like 3+ different courses/videos/tutorials
jacek: have you done xor example?
ImSeaFu: Like I get that you use data to create a functioon, I get that learning rate is how big of steps it's going to take, that kind of stuff but I have no idea how to implement any of this information
ImSeaFu: I was looking at https://www.codingame.com/playgrounds/59631/neural-network-xor-example-from-scratch-no-libs
Wontonimo: hey, i've heard of that author ;)
ImSeaFu: and I just got completely confused, I have no idea what it's trying to learn, I have no idea where it's getting the data from, I see a bunch of familiar things in the code but as a whole I don't get it xD
Astrobytes: you can check the version *with* libs that's out there
Astrobytes: before going to jaceks one
Wontonimo: do you know matrix math ImSeaFu ?
Wontonimo: like if X, M, and B are matries, could you compute MX + B
ImSeaFu: 5DN1L thought that it would be better to learn without the libraries so that I actually understand what the libraries are doing
ImSeaFu: Uhhh no
Astrobytes: Either way, you should look at them both.
Wontonimo: AH! okay man, that there is something you want to review if you want to do NNs from scratch
jacek: libraries do magic. you cant learn from magic much
Wontonimo: but ignore NNs.
Wontonimo: let's focus on ML like in my video
Astrobytes: Yeah but you see what they produce. Then go for the juicy stuff
Wontonimo: where you are trying to 'discover' values from data
Wontonimo: do you think you could write that code in the video without looking at the video or tthe links?
Astrobytes: Wait, no linear algebra experience at all or... ? ImSeaFu
jacek: he only done non-linear algebra ~
ImSeaFu: I'm very bad with terminology when it comes to math. Like I have experience with the graphing quadratic equations, and that kind of shizz but idk if that's the same stuff xD
ImSeaFu: wait which code? did I already forget? I remember you talking about checkers xD
Astrobytes: linear equations, transformations, vectors, matrices...
ImSeaFu: responding to wontonimo
ImSeaFu: but I think that answers the question xD
ImSeaFu: No I could not do that
Astrobytes: A basic foundation in linear algebra would be helpful to you I think, don't worry, it's not that bad
ImSeaFu: Lol maybe I'm too dumb for ML xD
ImSeaFu: Maybe someone can come up with a model that can train me? xD
Astrobytes: dumb is a state of mind
Astrobytes: Use your personal NN.
Astrobytes: If you find you are lacking in prerequisites for LA then go back and do those too. It will serve you well either way
ImSeaFu: Whoever put together my NN needs to go back to the drawing board xD
Astrobytes: Are you at college or school or ?
ImSeaFu: Due to mental health I wasn't able to go to post secondary
ImSeaFu: Now I'm idk like 32 or something
ImSeaFu: Trying to learn this and struggling quite thoroughly xD
Wontonimo: hey ImSeaFu, you said that you've seen that same content 3 times but that you can't code it without reference. From that I will recommend _practicing_ coding it from memory. The act of doing so will solidify the concepts in my mind
Astrobytes: That's cool, and still doable. And listen to Wontonimo.
Astrobytes: You are also trying to run before you crawl
Astrobytes: Which isn't a bad ambition but it's bad practice
Wontonimo: it's like trying to juggle, ride a unicycle, and play a harmonica all at once. It's not effective. Practice one small element (juggling) and perfect it. In this case, it's recalling doing parameter optimization like in the video by heart
Wontonimo: gotta go
Astrobytes: later Wontonimo, take care
Wontonimo: keep in touch ImSeaFu and let me know how you progress
ImSeaFu: I think what I'll do is write your code by hand and then try to comment it to see if I understand it
ImSeaFu: toodles wonton
Astrobytes: I'm out too, music and beer time, see you all tomorrow!
Astrobytes: (or whenever)
ImSeaFu: Lol I mean atm I've been up for something like 40 hour straight so I'm probably gonna wait til I've slept to bang my head at this some more
ImSeaFu: oh shhhht I'm not 32 I'm 31 lol it's not been my birthday yet
wlesavo: hey, in codebusters, when the ghost is standing on top of a buster it wont ever move? cant find this in referee
struct: Can I link you the referee wlesavo?
wlesavo: struct i found the eulers one, but i cant derive the bad behaviour from it
struct: im not sure if thats the one
struct: or the one that is in puzzle directory
wlesavo: https://www.codingame.com/replay/634789361 here, turn 145
wlesavo: oh, thx, this is different from the one i found
struct: This is the leaked one I think
struct: But I would check the puzzle one too, it might be different
PatrickMcGinnisII: line 643, i don't think ghost is release because there is only 1 ghost, I mean is not ... why there is a check to see if ghost id < ghost.size is beyond me
jacek: happy Caturday
antiwonto: [auto] Hey jacek, here is a :taco: for loggin in today while it is quiet. You now have 16 tacos