ASM_MOV_INT: i had to start over, so i took a break, when i realized on the fourth test, that the number are hex...! and I was up to the 50** commands too
One-Six: ohno......
ASM_MOV_INT: rofl
One-Six: I'm trying it out too but with Lua
One-Six: which doesn't play that nicely with binary and hex, xd
One-Six: I probably need to write a bunch of conversion functions, I'll see
ASM_MOV_INT: that puzzle started me on the function writing path in javascript...
ASM_MOV_INT: i will revisit it on another night, i'm at work, and only have an hour left
ASM_MOV_INT: :)
One-Six: For me, I'm trying out table of functions for Lua (table being a glorified array that can store pretty much anything)
One-Six: Ah, I see
ASM_MOV_INT: table of functions? sounds interesting
One-Six: yeah, so to read and perform a opcode I can do something like op[*first char of input*](input)
ASM_MOV_INT: I just have a getReg() and setReg() function, then just started churning through the commands one at a time... parsing commands is a very good programming exercise
One-Six: ah... I did it the bad way (in a way) and used a global table as the register and a global var as instruction counter
ASM_MOV_INT: well, now you got me looking at it again! rofl... oops i only passed test 1, test 2 contains the hex
One-Six: ohno
ASM_MOV_INT: :wink:
One-Six: and I decided to use a lookup table for hex2bin 'cause why not, XD
One-Six: now I just need to implement bitwise operations
ASM_MOV_INT: lookup table? nice
One-Six: wait, surely Lua have those...?
One-Six: Oh I think Lua 5.2 has a binary library, but I don't know about other versions
One-Six: I think 5.0 through 5.4 coexists and different version is used by different implementations... Lua is a pretty botchy language, but it's easy to get into for sure
One-Six: found it, 5.3 onwards natively support several bitwise operations w/o need for libraries.
ASM_MOV_INT: what version of lua is on here?
One-Six: I have to assume it's the newest (5.4)
One-Six: yep
One-Six: at least you can get the version by print(_VERSION), xd
ASM_MOV_INT: :grinning:
One-Six: So no more lookup tables, hex2dec2hex it is then
ASM_MOV_INT: oh wow, I found a bug in my bitwise OR function for opcode 40xy, fixed it... passed test 2
One-Six: yay
ASM_MOV_INT: this has motivated me to move forward
ASM_MOV_INT: rofl
One-Six: xd
One-Six: I ended up with another lookup table inside of the dec2hex function
One-Six: because of ABCDEF
One-Six: XDD
ASM_MOV_INT: I made a getNN function
One-Six: In my case, Lua had string.sub(str,i,j)
One-Six: so I condensed it to s(str,d1,d2=d1)
One-Six: so I would do s(opcode,3,4)
One-Six: (oh yeah I decided to just store all hex as strings and just convert them to intergers when I need to mess with them)
ASM_MOV_INT: omg i got it!
Dorak: Simple CPU Emulation?
ASM_MOV_INT: I thought it would take longer than that....
One-Six: whoa
One-Six: nice
ASM_MOV_INT: ty <bow>
ASM_MOV_INT: ooh it's a hard puzzle
ASM_MOV_INT: hmmmm
ASM_MOV_INT: so that one guy earlier that said he only got 50%
Dorak: Was me. xD
ASM_MOV_INT: if the program doesn't include the 0000 at the end, don't print
One-Six: That would be... yeah, Dorak
One-Six: Ohhh...
struct: yes its lua 5.4.3
One-Six: Thanks for verifying, struct!
ASM_MOV_INT: at least that was my first thought after seeing his comments on chat, then reading the problem... first thing I implemented was to put that in a condition only if it was a command
Dorak: I managed to do it a while ago. I understood my mistake, it was a base error.
ASM_MOV_INT: ah, yes the hex
ASM_MOV_INT: i blame the hex
ASM_MOV_INT: rofl
One-Six: I'm back
One-Six: now let's see if I know how bitwise operations work...
**Tien2k9 slaps One-Six around a bit with a large fishbot
struct: what are you workign on?
One-Six: Simple CPU Emulation 1, what ASM and Dorak had been working on
One-Six: huh, it didn't mention how to deal with subtraction overflow – or addition overflow clearly for that matter
One-Six: let me pretend subtraction never overflows for now xd
Tien2k9: (╯°□°）╯︵ ┻━┻
One-Six: oh, I should pretend it uses two's compliment...?
Tien2k9: yay i win clash 3 times a row :D
One-Six: nice
One-Six: I've been doing clash just for the heck of it, but I do get first and seconds so that's real nice
One-Six: always beat by python and ruby in shortest mode though XD
One-Six: shoot, my intricate code returns nothing XD
One-Six: nice I did it
One-Six: I think
FoXbOw: https://www.codingame.com/clashofcode/clash/22055844a44cd27c0c856361cb2548721723cc6
One-Six: is there a way to reopen a CoC channel I exited?
ASM_MOV_INT: yes, but it will take you back to the puzzle, not the actual clash
ASM_MOV_INT: ok, i had to get it...
ASM_MOV_INT: https://eulerschezahl.herokuapp.com/codingame/puzzles/
One-Six: Oh, I mean the chat-channel – I think it's called chat room?
One-Six: whoa, that's a cool website
One-Six: Nice! I aced that CPU puzzle!
ASM_MOV_INT: not sure about the chat rooms
ASM_MOV_INT: nice
ASM_MOV_INT: very cool puzzle
One-Six: It says I am should've learned about stacks with it
One-Six: I don't think I used stacks at all? XD
One-Six: Just chucked the entire raw opcode into a table and used a separate instruction index
One-Six: Let's see if there's CPU Emulation 2
One-Six: Uh, probably not tonight though (
One-Six: noooo there isn't
One-Six: Oh I know, I can write one! ((
One-Six: oh, there's already some brainf**k themed puzzles
ASM_MOV_INT: I didn't use stacks either....
One-Six: y'all should try brainf**k, it's super fun to mess with XD
ASM_MOV_INT: where at? link
One-Six: en.wikipedia.org/wiki/Brainf**k
One-Six: Mitxela also made this super nice interpreter with visible memories:
One-Six: https://mitxela.com/other/brainfuck
One-Six: lemme find something I wrote a long time ago
ASM_MOV_INT: rofl, and it passed the censor test lol
One-Six: guess it's given a special case XDD
derjack: :upside_down:
One-Six: Here's something I cooked up:
One-Six: https://pastebin.com/4c34Zir8
One-Six: You will also need this extended ASCII character for it to work: ÿ
One-Six: (because I used FF for EOF)
One-Six: (I don't remember how each bit works exactly now though, but looking at the memory, you should still be able to tell what each bit does – kind of?
One-Six: oh, mixtela also have a YT video explaining how BF works and how he made a TicTacToe "A.I." with it
One-Six: that man is rad
Dorak: What version of net core (or net framework) does the page use? It gives me a thousand errors a valid F# solution.
struct: .NET Core Runtime 3.1.3
struct: for F#
Dorak: Uhh... Thanks.
Dorak: For C# is net core 3.1 too?
struct: yes
struct: C# 8.0 (.NET Core Runtime 3.1.3)
Compiled with ‑unsafe
Dorak: Does it take a long time for the page to update the language versions? We are currently on NET Core 6 (lts version). The new C# and F# features since .net 5+ would be really useful here. :cry:
struct: They usually update them before contests start
Dorak: Just out of curiosity, what contests are you talking about? Events like spring challenge 2022?
struct: yeah
struct: I dont know when the next one is
Dorak: Okok. Thank you very much for your time.
struct: np
Dorak: Just one more question. Is the information on the languages and their versions public? I would like to see info about other languages too.
struct: https://www.codingame.com/playgrounds/40701/help-center/languages-versions
Dorak: Thanks a lot. :kissing_closed_eyes:
struct: It doesnt get update often due to backwards compatibility I think
struct: Some languages updates can break some bots
ashiknur: can anyone suggest to me a book on algorithms needs in codingame? l
ashiknur: like minimax, monte charlo, etc
struct: I dont know any sorry, but there are plenty of resources online for those 2 algos
struct: are you mostly interested in AI?
derjack: no books, just random articles here and there
derjack: and/or chess programming wiki
struct: no books for NN aswell?
ashiknur: @struct I want to learn those algorithms actually needs for game bot. I have participated on many codingame bot competition without knowing any AI algorithms. btw I managed to get to silver in some of those. so I want to get it to a new level.
derjack: im not a books guy. but apparently the one by sutton is good
struct: yeap Im also not a book fan
struct: ashiknur I would pick a simple game to play and try to implement those algos there
struct: for example connect 4 is easy to tell whats happening, the sim is trivial
derjack: minimax video https://www.youtube.com/watch?v=KU9Ch59-4vw
derjack: well UTTT wood league is simple tic tac toe
ashiknur: okay, then suggest me some resources. it would be great If I find a list of algorithms needs in game bot programming
struct: monte carlo (random search), genetic algorithm, beam search, minimax, mcts
struct: Those are probably the most used ones
ashiknur: I learned minimax, genetic algorithm, neural network ( still stuck on backpropagation) know the algorithm mcts didn't find a good resource for implementing part. I am c++ programmer. I know only c++ language.
derjack: general: implement game simulation: list all moves, make move, undo move, check for end conditions, who won etc.
understand minimax, then alpha-beta which is faster than minimax, then iterative deepening to control for time
understand monte carlo, random simulations, then UCB1 formula for monte carlo tree search
others: genetic algorithm, beam search, bfs/dfs
derjack: NN simple example https://www.codingame.com/playgrounds/59631/neural-network-xor-example-from-scratch-no-libs
derjack: in python, but should be easily translatable to c++
ashiknur: okay, I will surely work on this implementation.
struct: bitboards are also important if you are into board games
struct: but one thing at a time
ashiknur: I didn't hear about that bitboards. thanks for suggesting this. I am really interested in board games.
derjack: yeah they are generaly at least order of magnitude faster than 'normal' approaches
struct: and easier to code in once you finally understand them
Big.Tom: JL727
Big.Tom: SUIIIIIIIII
Big.Tom: LAKAKA
Big.Tom: EMILE SMITH ROWE
Big.Tom: SUIIIIIIIIIIIIIIII
5DN1L: No spam please, English only please
JohntheHacker: Know amost nothing about coding
JohntheHacker: *almost
yuutz: suiiiiiii
JohntheHacker: ???
Minimaliist: suiiiiii
Big.Tom: HEHEJE SUIIIII
5DN1L: This site is for practising, not for you to learn from nothing, JohntheHacker
JohntheHacker: Where can I learn javascript?
5DN1L: Google
JohntheHacker: ok
JohntheHacker: thanks
5DN1L: You're welcome
orchidperson: gg
JCMolhant: Hello people, I'm stuck on the Thor game but I'm pretty sure my code is sound
JCMolhant: Anybody willing to check it for me ?
JCMolhant: :disappointed:
bonsaifree: language?
JCMolhant: javascript
bonsaifree: im not into js nowadays, but it should be an easy, so you can send pm if you want
JCMolhant: Thanks ! Doing just that now
BadBurger: wait, theres not marslander 2? if i remember there was one couple years ago?
BadBurger: yes there is, just didnt find it in search
Rebollo: Guys I lost the code u sent these days
Rebollo: How can I convert
Rebollo: int
Rebollo: to Enum Int
struct: static_cast<a>()?
5DN1L: Rebeollo: And you may try searching in the chat log: https://cg.spdns.eu/wiki/Special:PrefixIndex?prefix=World%2F&namespace=3000
5DN1L: Rebollo*
Rebollo: thanks
One-Six: oh my god, there's a chat log hosted on mediawiki?
One-Six: ...There must be a better way XD
One-Six: Can I help with the chat log formatting? I'm reasonably familiar in Wikitext
5DN1L: It's not official
5DN1L: I don't know who created the page in the first place :thinking:
derjack: dbdr?
One-Six: :thinking:
5DN1L: Should One-Six ask dbdr?
derjack: or ask Automaton2000
Automaton2000: for me is that i was supposed to be a moderator
struct: i guess you could try messaging dbdr on the forum, im not sure if he is online often
derjack: dbdr seems online now
Banble: why can you not clash with html?
solution.cpp: because html is not a programming language
Rebollo: Guys when I get segmentation fault, is there a way to interpret the messa errorr?
Rebollo: Segmentation fault.
at Answer.cpp. function golfGame::addNode (this=0x7fffffffd630, obj=..., i=0, newLevel=true, d=0) on line 237
at Answer.cpp. function golfGame::loadTree (this=0x7fffffffd630) on line 447
at Answer.cpp. function main () on line 487
struct: best way is to run the code on an ide
DaNinja: node might be null when you call node.addNode()
CryanRyan: why would they add a world chat and not allow spamming, I'm not even on the discord
darkhorse64: rebollo: line 237 is where the error occurs. You are trying to access an invalid memory address either because you have a null pointer or you are accessing a array out of its bounds
struct: these errors can be tricky to figure out
darkhorse64: definitely. If possible, use an IDE to debug because using cerr to find problems can be difficult
Rebollo: which one you guys suggest?
struct: are you on windows?
Rebollo: yes
struct: I suggest visual studio
struct: but it uses a different compiler that cg uses
struct: but should be fine for most of the stuff
struct: its a bit heavy though
darkhorse64: ^
Rebollo: do you guys see aanything wrong with this function/
Rebollo: http://chat.codingame.com/pastebin/e30a38fc-3ca4-4b2c-afec-e3de69e9861d
Rebollo: Does the pointer work only inside of it? And get lost after it is returned?
Rebollo: Also I am intilizaling the childs pointer as nullpointer
Rebollo: is it possible to aassign after I instatiate the object?
struct: that code seems fine
struct: could be reduced to
struct: objectsNode *createNode(objects obj) {
return new objectsNode(obj, nullptr, nullptr, nullptr, nullptr);
}
struct: Yes you can assign a pointer after
Rebollo: But once I am creating it inside the function, is is going to exist after it as well/
darkhorse64: With the new VS 2022, you can even compile for the Linux subsystem. This way, you are making sure you are generating code similar to what runs on CG servers. I have noticed that floating point computations behave differently between msvc and gcc
struct: yes Rebollo
darkhorse64: Memory allocation is persistent
Rebollo: I am loosing the hope
struct: Rebollo how long have you been coding for?
struct: I think I asked you this before
struct: but cant remember
darkhorse64: Your code is correct but you are allocating node by node while computing which is very bad for performance. Allocate a pool of nodes at startup
Rebollo: c++ just for months
Rebollo: but I work with SQL, data base
Rebollo: and other laguages for 15 yearsa
struct: These are normal errors for c++ programers
struct: dont stress about it
struct: its probably out of bounds error like was said before
therealbeef: I remember you are learning, but I would advice avoiding the use of new and raw pointers
Rebollo: I am trying to create a pointer, like and empty NODE, just to behave as a structure for the tree
Rebollo: the problem is in the following commented row
Rebollo: http://chat.codingame.com/pastebin/d48afaba-a60c-4862-a2bc-dcabd54d5501
Rebollo: It is a vector of pointers
Rebollo: that I am trying to use as many roots for the tree
struct: are you sure i is within the current size of the vector?
Rebollo: But once I am using a vector isnt it just a matter of assign any value to the index and it going to create the position?
struct: no
Rebollo: vector[1]
Rebollo: vector[1]
Rebollo: vector[1]
struct: vector<int> vec;
vec[0] = 5;
struct: that code wont work
struct: vector<int> vec(1);
vec[0] = 5;
struct: works
struct: if you want to add a new element
struct: you can just do push_back or emplace_back
Rebollo: hum
Rebollo: its size is 0
DaNinja: headNode.push_back(tempLevel)
Rebollo: can I copy from ohter vector?
Rebollo: headNode = ballNode
struct: if you can copy a vector into another?
Rebollo: like just assigning
Rebollo: vector1OfPointers = Vector2OfPoisnters
struct: yes
Rebollo: yes it worked now
Rebollo: :lips:
Rebollo: sorry guys
Rebollo: filling a idiot
Rebollo: but thats ok
Rebollo: laways learning
Rebollo: I did that:
Rebollo: objectsNode *tempNode = ballNodes[i];
Rebollo: Now every change in the first pointer will reflect in the other?
Rebollo: Like assigning childs and so on?
Rebollo: HUm it will
Rebollo: worked
TheModified: i just won my first full lobby clash!
Squar3: it will happen a lot of times just wait for it ^^
mrswagger21: congratsss
Squar3: can I understand why my code isn't optimized for all of the medium puzzles :( ?
Squar3: I have 3 70% because of optimization
5DN1L: which puzzle
Squar3: Stock Exchange Losses
and
Remaining Card
5DN1L: both 70%?
Squar3: one 70% and one 60% both only optimzation left
Squar3: and there is one more task but m working on it
5DN1L: oh, for Remaining Card, I just look up the relevant number sequence...
Squar3: I have no idea what that means but I'll search for it
5DN1L: and for Stock Exchange Losses, you're supposed to read through the numbers once and adjust your answer while you're reading it
5DN1L: much like the Temperatures puzzle if you do it that way
Squar3: yep I did the temperatures ones
WhatInTheChap: giz hos do progrem??
5DN1L: English please
WhatInTheChap: poo
WhatInTheChap: hi*
5DN1L: Reminder, in case you forget, https://www.codingame.com/playgrounds/40701/help-center/code-of-conduct
WhatInTheChap: 69
WhatInTheChap: lol!!!!!!!!!!!!!!
5DN1L: you want to spam, then i want to kick you out
WhatInTheChap: python question
Fasader: good day fellow coders!
WhatInTheChap: Good day!
Fasader: state thine query
Squar3: bro I finished the stock exchange finally xD
5DN1L: good job
Squar3: I found another way of doing it
Squar3: my first way was too dumb
5DN1L: :D
Squar3: damn I thought the problem is that m bad at C lang xD
WhatInTheChap: How do you output in programming?!
5DN1L: read the default code
5DN1L: it tells you how
Squar3: yup I used the same way as solution 2
WhatInTheChap: beanz
Squar3: 5DN1L is the puzzle of the week always "hard" diffuclity ?
5DN1L: no
5DN1L: it can be in any difficulty
Squar3: voted based one what then?
Squar3: how much ppl think it's interesting?
5DN1L: based on CG staff's choice
Squar3: oh man I just saw your profile.. u r fking amazing
5DN1L: :blush:
Squar3: how r u good with all of these programming langauges ????????? xD
5DN1L: i'm not good... i just finish enough number of them lol
Squar3: haha still idk how u remember the syntax lmao
5DN1L: i don't
5DN1L: i just google
Squar3: ah I get it
Fasader: WhatInTheChap just look at the hello world code and see how it's done. in python its "print()" but in other languages it's usually something different
Dorak: Could any mod tell me how to unlock the "optimization" path. :thinking:
5DN1L: the graph tells you how?
Wontonimo: you need to progress in AI or Algorithms
AntiSquid: https://gab.com/WesternChauvinist1/posts/107726274007953968
Dorak: Thanks wontonimo.
Wontonimo: that guy has a point. those protesters are stupid and just harassing the public
Wontonimo: hey Dorak, you need 2 achievements in either of those to unlock the path. Do yo have either and if so what achievement?
Dorak: I still haven't reached the bronze league in two bots, I only have one in silver. And I still have to do the puzzle of the week, I imagine that I must complete at least one of those two achievements.
Ftyghi: hi
Ftyghi: does anyone remeber me
infox: no
BadBurger: need some help, in kotlin, using swing to draw a image for marslander sim, can i modify the return object to alter the g2.color for example without re assigning the class again?
BadBurger: http://chat.codingame.com/pastebin/6907ee1c-8aa2-4eda-837c-641a4b9d9bd9
Ftyghi: infox not youWontonimo
Wontonimo: hey Ftyghi, yeah I remember you
Ftyghi: hi
Ftyghi: dude what your time
Ftyghi: i have to go
Wontonimo: 9:18am
Wontonimo: bye
Ftyghi: bye
Ftyghi: mine is 5 20 pm
Ftyghi: bye
darkhorse64: There is a new optim in the contributions based on the travelling salesman problem. It looks good to me but it would be nice if more people review it
theeguyhoocarntspel: Hog Ridaaaaa
Lambert_W_Function: hooooooooog ridaaaaaaaaa
Lambert_W_Function: please dont mute me
5DN1L: :smirk:
5DN1L: next time I will, even if you say it
CommanderAzure: why say that
USSR_Chicken: hello
Quidome: hi
Depki7: hiho
USSR_Chicken: so, im stuck on the puzzle "The decsent"
5DN1L: You've achieved 100% already?
USSR_Chicken: it said good job and told me to submit it even though i didnt hit anything
USSR_Chicken: http://chat.codingame.com/pastebin/12349d7b-bd3b-4ae5-9824-723afa067236
USSR_Chicken: heh
USSR_Chicken: oops
USSR_Chicken: i guess we can try a different one then
USSR_Chicken: imma try power of thor
USSR_Chicken: oh no my teacher is coming
Astrobytes: pay attention in school, it's important
Wontonimo: you must be kidding Astrobytes. A whole country collecting taxes, building schools, hiring teachers, paying for supplies, taking attendance, coming up with lesson plans, standardize testing to try and improve learning. Nah, it's probably not important, just a big conspiracy to waste everyones time and money
5DN1L: all the generalisations... anyway, let's get back to coding
Miki09: What means "All inputs and outputs are mandatory" in exitor?
Atomcast: hey guys
Emmus: hello
jacek: :upside_down:
ninjadip: loved the docu "waiting for Superman" it's on America's school system
ninjadip: or failing school system
ninjadip: how more bigger buddgets don't make smarter kids
ninjadip: how bigger...*
5DN1L: aren't you a smart kid? :D
jacek: oO
5DN1L: ok, jacek is a smart kid too :D
situu: Hello!
Ose: hello
EzS-NoVa: https://www.codingame.com/clashofcode/clash/2206733beeed73fef60a1b5f4e8fe78b0f2e5be
EzS-NoVa: somebody wanna play in reverse?
5DN1L: please use #clash channel for clash invites, thanks.
jacek: if only we would have bot for that...
5DN1L: yup
Wontonimo: just about to deploy it
Blokops: is it Automaton2000 ?
Automaton2000: so there is a real problem
Miki09: Why codingame has maximum lenght of output in contributions?
Miki09: I'm currently making new testcases to my contrib and I have reached the max number of characters in my big testcase :(
jacek: oO
Wontonimo: okay, a trial run of the clash of code automated message is now running. if someone posts a clash link , I'll auto respond in 4 sec
Miki09: lets try
Blokops: come join my clash ~ ~ ~
Blokops: https://www.codingame.com/clashofcode/clash/2206733beeed73fef60a1b5f4e8fe78b0f2e5be
wontonimo: [automated] hey Blokops, dont paste those links here. Use the channel #clash
Blokops: yup is working
Wontonimo: cool! it worked
Husoski: https://www.codingame.com/clashofcode/clash/report/2206774f26dbc3e5f642b8b90721d2449359840
wontonimo: [automated] hey Husoski, dont paste those links here. Use the channel #clash
Blokops: Automaton2000 they taking over your job
Automaton2000: madknight why are you guys in?
Husoski: That caught a "results" page...isn't showing that okay?
Adam.e: big body benz
Blokops: https://www.codingame.com/clashofcode/clash
wontonimo: [automated] hey Blokops, dont paste those links here. Use the channel #clash
Adam.e: whos got the biggest bunda
Adam.e: !
Blokops: https://www.codingame.com/clashofcode
Wontonimo: that's enough Blokops
Adam.e: ??
Adam.e: shush
Adam.e: dumb yute
Blokops: just edge testing
Adam.e: edge testing?
Adam.e: have you been to eez foods
Adam.e: they sell deez nuts
Adam.e: !!!!!!
5DN1L: don't spam
Wontonimo: no more spamming Adam.e
Adam.e: who gon stop me
Wontonimo: i'm giving you a 5 min time out to cool off.
POEJordanJ: rip lmao
5DN1L: Do you have to be online for the bot to work, Wontonimo?
Wontonimo: ope
Wontonimo: nope
5DN1L: cool
Wontonimo: just my server has to be on
struct: :sleepy:
Husoski: Wontonimo, consider filtering on "/clashofcode/clash/", *not* followed by "report". It should be okay to post a result link if someone's asking a question about a clash, right?
5DN1L: do they need to post a link to the report?
Blokops: ^ they ca just post the code
Husoski: Wah! Why did I come in 2nd on https://www.codingame.com/clashofcode/clash/report/2206774f26dbc3e5f642b8b90721d2449359840
wontonimo: [automated] hey Husoski, dont paste those links here. Use the channel #clash
5DN1L: i don't know
5DN1L: pass rate then time?
5DN1L: i think you have a point there
struct: Wontonimo from 1 to 10 how important are derivatives for NNs?
5DN1L: but again
5DN1L: we can answer it without the report
Husoski: yes, of course, but it still got the direction to #clash, where it probably won't get any help
jay3332: te
Wontonimo: hey Husoski (and everryone) you can test the bot responses in #wontotest
5DN1L: it's ok either way i guess
Wontonimo: i added your suggestion
Wontonimo: struct, what do you mean about derivatives ?
Wontonimo: like, how important is it to understand them
Husoski: #wontotest
Wontonimo: or do you need them to make NNs work
struct: how important is it to understand them
5DN1L: Husoski true
Wontonimo: you have to click the link Husoski or type /join wontotest
Husoski: thx
Wontonimo: you don't really need to understand them
struct: i only slept 3 hours today and reached that part of the book and didnt understand it very well
struct: i probably will reread it tomorrow
jacek: you only need to know that for backprop for some reason you need to multiply by derivative of activation function
Wontonimo: but knowing that derivative is just "the slope" is good
Wontonimo: knowing it is the slop is important
Wontonimo: let's say you have some function f(x) = a*x + b
Wontonimo: and you have tried a particular value of A and B and you get some error 11.2
Wontonimo: and you pick a value really close to A, say A+0.001
Wontonimo: and the error is 11.204
Wontonimo: the would could say that the slope is .004 error / .001 A
Wontonimo: or 4 error / 1 A
Wontonimo: now that we know the direction that the error will change and by how much in relationship to A, we cold say that we want to try and get rid of all the error
Wontonimo: we have 11.2 error ... so if we change A by - 11.2/4 = -2.8
Wontonimo: then that is equal to a "learning rate of 1". We use the FULL error correction
Wontonimo: a learning rate of 0.1 would mean that we would only use -2.8*0.1 = -0.28
Wontonimo: we would only correct A a bit, moving closer to the estimated full error correction
Wontonimo: does that make sense or did I just confuse the issue?
struct: Makes sense yes
struct: I will reread it though, I feel like its important for backprop
Wontonimo: yeah, but it's just "slope"
struct: yeah
Pollard_Rho_Algorithm: hello everyone
Pollard_Rho_Algorithm: it is me, pollard_rho_algorithm
struct: slope of the tangent right?
Astrobytes: yes
struct: thanks
Astrobytes: or 'rate of change'
struct: soon my bt will be finished and I can finally rest
Astrobytes: as wontonimo said you can find the rate of change in whatever direction minimises the error, then you move in that direction. I think that's part of gradient descent
Astrobytes: been a minute since I did any proper calculus
Wontonimo: yes, that is a great description Astrobytes
struct: yeah, im not great at calculus
struct: Im probably just tired
Astrobytes: yeah, get some sleep, it's not overly complex in terms of concept, just lots of symbols
struct: brb
Blokops: is there a way to see the game code for the contest
Blokops: not the answer but the game itself
Wontonimo: yeah
Wontonimo: i'll dm you
Wontonimo: oh no dbdr took down his github site!
Wontonimo: anyone have a fork of brutal tester?
Wontonimo: oh, nevermind. looks like it wasn't dbdr. it's all of github
Blokops: it would be apocalyptic if github lost all its data
Blokops: 2k panic levels of disaster
mrswagger21: I lost the game :sob:
jacek: oO
jacek: happy groundhog day, again
supertebibyte: is there a way to replay/finish your clashes?
supertebibyte: i can't find the puzzle or my code once i have submitted it
jacek: you can use https://eulerschezahl.herokuapp.com/codingame/puzzles/ to search for it
supertebibyte: thank you, i was able to find one of them. the reverse ones are harder to find
jacek: thats the point of the reverse
Blokops: couldn't someone just have the answer before hand using that page?
supertebibyte: yes, it defeats the point if used maliciously. but i just want to finish my "thought" as it were, when the 10min timer is over.
jacek: actually you have 15 minutes
supertebibyte: either way :)
supertebibyte: i guess i'll just hope it comes up again some time
Jayarrowz: :grin:
jacek: 1 2 7 3...
Idk.-_-: hi
Idk.-_-: guys guys
jacek: good evening
Jayarrowz: Reverse is a bit funny
Jayarrowz: sometimes
jacek: yhw
Fasader: ynnuf tib a si esreveR
Idk.-_-: do you think wich is the best and best look im say best proggraming language is who
Idk.-_-: fastest esaiest bestest language
One-Six: There isn't really a "best programming language" if that's what you are asking...?
Fasader: what are you even saying?
One-Six: Oh, easiest to learn might be Python?
One-Six: Lua is pretty easy to get into but can get messy if you want to do advanced stuff
One-Six: Python seems to be well received and extremely useful almost everywhere
jacek: easiest would to just tell the computer what you want and it will do it
Idk.-_-: :D
One-Six: Ah, yes, the Alexa programming language
One-Six: "It can wake you up!"
Idk.-_-: no it's jarvis
jacek: it would be quite distant(?) future, but its doaeble
One-Six: natural languages are inherently ambiguous though
One-Six: translating that into non ambiguous code can be hard for both the translator and the dictator
Idk.-_-: and 1 q. which language is best after python ?
One-Six: idk XD
One-Six: want to plug Lua, but Lua starts counting at 1 so...
Idk.-_-: mate you can't just say the simple answer ?
jacek: the question is more opinion-based i think
jacek: and python or ruby is quite than any other language
jacek: its a peanut butter jelly time!
5DN1L: pascal basic julia time?
jacek: Automaton2000 do you like peanut butter?
Automaton2000: i guess you can add it to the right of (0, 0). expected (2, 0).
jacek: :x
erinc--emre: Shortest Mode is ON
erinc--emre: https://www.codingame.com/clashofcode/clash/2207154cef8923aa72979f5780b15af8406d1a8
wontonimo: [automated] hey erinc--emre, dont paste those links here. Use the channel #clash
jacek: oh my
BadBurger: is atan2 same is radians(thetan)?
Fasader: atan is just one letter away from satan
Zbouig: baguette
Zbouig: froggies
Uljahn: Automaton2000 do you like peanut butter on pizza?
Automaton2000: and that is the one that you can only make one move per turn?
Wontonimo: woot! my autobot responded to a clash post!
Papi.Chulo: :eggplant:
MessitheGoAt1910: :sweat_drops:
Papi.Chulo: :peach:
Husoski: BadBurger No. atan2(y,x) gives the angle (yes, in radians) between the ray from (0,0) to (x,y) and the +x axis.
Husoski: radians(th) converts degrees to radians.
Papi.Chulo: chupa mi pene porfa
I.T.: nudes?
Papi.Chulo: si
Wontonimo: Husoski, you want any help with mad pod racing?
Wontonimo: you are level 33! come join the under 2000 ranking
HUNTERxHUNTER: :flushed:
Husoski: Wontonimo Can I take you up on that offer later? (But I may want some more optimization hints first, re Marslander 2)
Wontonimo: yeah
Wontonimo: you want some hints for mars 2?
Wontonimo: have you looked at PID controllers? https://w3.cs.jmu.edu/spragunr/CS354_F17/handouts/pid.pdf
HUNTERxHUNTER: penis?
Wontonimo: HUNTERxHUNTER you have been given a 10 min time out. please feel free to come back but be respectful
Husoski: Yeah, we talked about that. What I see is a method for computing an input change based on estimate of an error term, but I don't get how to estimate an error for a trajectory when I don't know what that trajectory should be.
Husoski: ...and I have that PDF on my desktop (but it took some fiddling to get it into a form that my ancient Acrobat would understand.)
Wontonimo: so, imagine we are in a car going towards red light. We don't hit the breaks when we are at the red light. There is some distance given our speed where we should slow down
Wontonimo: if our distance is 100m
Wontonimo: and our speed is 10m/s
Wontonimo: then we will blow through that red light in 10sec
Husoski: That's fine unitl mountans get in the way.
Wontonimo: haha
Wontonimo: let's say we can change our speed by 2m/s*s , in other words we can reduce our speed by 2m/s every second
Husoski: yes...
Wontonimo: so if we say, hey if we are going to blow through that light in the next 6 sec, then slow down. else everything is fine
Husoski: (but it's really about 0.3 m/s^2 for Marslander)
Wontonimo: so, let's see how that plays out
Wontonimo: 0 sec , dist = 100 , vel = 10 , eta =10
Wontonimo: therefore no breaking (eta > 6)
Wontonimo: 1 sec , dist = 90, vel = 10 , eta = 9
Wontonimo: 2 sec , dist = 80 , vel = 10 , eta = 8
Wontonimo: 6 sec , dist = 60 , vel = 10 , eta = 6
slow down !
Wontonimo: 7 sec , dist = 52 , vel = 8 , eta = 6.5
no breaking
Wontonimo: 8 sec , dist = 44 , vel = 8 , eta = 5.5
slow down !
Husoski: I finished the lower division physics sequence, if that helps...
Wontonimo: and so on
Wontonimo: you can instead target the old location minus your speed correction
Husoski: So, how does this apply to the 2d trajectory problem? I need to reverse speed (later test/validators) and stay above the mountains and on the map while doing that...THEN figure out how to get back to flat ground and stay above the mountains.
Wontonimo: so,
new_target = target - 6 * velocity;
Wontonimo: if 6 is the correction factor
Husoski: where does that come from?
Wontonimo: that can translate directly into 2d
Wontonimo: from above
Husoski: 6?
Wontonimo: the factor (6) is found by experimentation usually, but if you happen to know the physics involved precisely you can derive a good guess
Husoski: I'm not getting a warm fuzzy feeling yet, but I'll head-scratch some more. Thanks!
Wontonimo: where the bot is aiming at the target location https://www.codingame.com/replay/575327101
and here it aims for target location - velocity
https://www.codingame.com/share-replay/575327237
Wontonimo: well, aims for location - 30*velocity
Wontonimo: and otherwise it is EXACTLY the same code
Husoski: And 30 was determined from the terrain, or is is hardcoded?
Husoski: Does that work on, say, test case 04 (deep canyon)?
Wontonimo: yeah, 30 was determined experimentally to work in a wide variety of situations
Wontonimo: given that I also constrain the banking (angle) to be fairly small
Wontonimo: the other bit of logic is what I call the "Big V"
Husoski: Big V?
Wontonimo: I project a large "V" from the landing pad, 45 degrees on both sides
Wontonimo: if the mars lander is outside the "Big V" then i use a different set of logic then inside
Wontonimo: when outside, the lander is not allowed to lose altitude
Wontonimo: when inside it will try and land
Husoski: That's a really useful idea.
Wontonimo: that's it.
Wontonimo: i don't keep track of mountains
Wontonimo: 45 degrees was just laziness , because that is y=x and y=-x
Husoski: I recall trying to not lose altitude while reversing velocity and ran off the side of the map, though...
ASM_MOV_INT: boy the syntax of programming languages that allow for such cryptic solutions....! phew
Husoski: asm_mov_int you can do all of this in C
Husoski: Oh, well, that gives me some stuff to try. Thanks again!
Wontonimo: right, if you limit angle then you also need to limit max sideways velocity
Wontonimo: or increase the coefficient to be larger than 30
Wontonimo: just because 30 worked for me, doesn't mean it will worrk for you.
Wontonimo: you have to experiment with a good factor for your particular code
ASM_MOV_INT: even assembly language is a bit more readable than most of these cryptic one liner solutions I see people churning out for clashes
Husoski: Oh, on your Markov puzzle I named my random generator class WontoRandom in your honor...
ASM_MOV_INT: :grinning:
Husoski: Yep. I was polluted by starting with a language called APL that *really* encouraged this sort of thing.
ASM_MOV_INT: interesting
Husoski: Here's a "complete" solution for an n-dimensional polynomial best fit in APL:
⌷⌹⌷○.*0,⍳⍴⌷
ASM_MOV_INT: i'm sure it is...!
ASM_MOV_INT: surprised i've never run across that APL language before, sounds right up my alley...
Husoski: You can still find it. Dyalog sells a commercial interpreter. If you answer a lot of questions right, you can get a free version for personal use.
Wontonimo: lol, thanks Husoski for the honor !
Husoski: Docs are slim and not so easy to find. The only data structure is an array. A bit like programming with numpy in Python.
Husoski: If you look at the docs for Python's itertools module in the standard library, you'll see that a lot of the standard iterator patterns are inspired by APL operators.
Husoski: Wontonimo My pleasure. The source is published. I only implemented the .choice() method, though, plus a .next() method in case I needed to debug something. I don't think it will pass the poker tests, though...
Husoski: :)
Husoski: It's dinner time on the US west coast. Type at you later...
ASM_MOV_INT: hey that's where I am!
Husoski: Correction: No rho in that APL polyfit code. Bad memory on my part, but it's 1 character shorter:
⌷⌹⌷○.*0,⍳⌷
Wontonimo: haha
Wontonimo: that's some crazy code
ASM_MOV_INT: sounds like a very efficient programming language
ashiknur: can anyone suggest to me a book on algorithms needs in codingame? l
