Chat:World/2022-01-23

From CG community
Revision as of 05:29, 24 January 2022 by Chat Log (talk | contribs) (Created page with "<img src=/a/77451004630011> Strong_Dog: hello File:Default_avatar.png orchidperson: hello <img src=/a/62842162200886> Smelty: hi <img src=/a/67353989164707> Mortis_666:...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

Strong_Dog: hello

Default avatar.png orchidperson: hello

Smelty: hi

Mortis_666: hi

Default avatar.png 1shadowless1: hi

jacek: c-c-c-co...

Default avatar.png Naknik: guys

Default avatar.png Naknik: how do I remove the auto \n in the c#

BlaiseEbuth: https://lmgt.org/?q=c%23+print+without+new+line

Default avatar.png JDone: Hi everyone, hope u're fine, I'm looking for a great way to learn about machine learning algorithms, does anyone has any class to suggest to me ?

MSmits: dont know about a class, but the best place to start is probably building the XOR example from scratch

MSmits: there's several websites and youtube tutorials

Default avatar.png JDone: thanks, I'm gonna try !

MSmits: don't start with tensorflow + keras immediately

MSmits: then you dont know what actually happens

MSmits: start using it when you know what the functions from those packages do

MSmits: this is obvious for any library or package, but when you import math in python, you kinda know what it does, but machine learning libraries can hide A LOT of important details in their functions

MSmits: after you do the xor, it might be fun to do MNIST

MSmits: there's a book that goes through that pretty handily, but i forget the title of it

Default avatar.png JDone: Yeah I heard about other libraries like ML5,... but the most important part is to understand the process I guess, more than the functions. If you remember the name of the book let me know :)

MSmits: someone here will know

MSmits: it's a book from a guy who does tons of ml videos

MSmits: I think this: https://www.youtube.com/watch?v=aircAruvnKk

MSmits: If you go on there, follow the videos and the links below it, you should get to tons of resources meant for someone like you who is just starting

therealbeef: There's always the 'bible' of AI, "AI a modern approach" from Russell and Norvig

MSmits: sounds good

therealbeef: It's the classic starting textbook of AI majors

MSmits: if you start with ML, i guess you'll skip a big part of that book right?

therealbeef: bit thick though

MSmits: like search algorithms/pathfinding etc.?

MSmits: bayesian probability

MSmits: inference

jacek: xor example from scratch?

therealbeef: Still interesting subjects to learn about ;) but you're right

jacek: where could you find one? :thinking:

MSmits: you wrote this out didnt you?

MSmits: i never read it because i had already done it with your advice before that

jacek: https://www.codingame.com/playgrounds/59631/neural-network-xor-example-from-scratch-no-libs

MSmits: right JDone, do that

jacek: and its free

MSmits: ye first time is free

jacek: and nor example is only $4.20

MSmits: how much is the nand?

jacek: $21.37

MSmits: ouch I'll just train separate networks for inverting + and

MSmits: btw, my oware bot plays perfectly in seeds 48-37 and 9-0 now, rank 10 is the best I can get without training the NN. Started working on it today, but will be a while. No more gains from meta mcts possible

MSmits: (unless i cheat and start adding stuff from my endgame book to counter book players)

jacek: i lost to nonNN :s https://www.codingame.com/replay/604377269

MSmits: i seriously doubt that's a non-NN

jacek: huh

MSmits: my bot was miles ahead of any non-NN and it's not gotten any weaker

MSmits: Renard is very strong

MSmits: (for a non-NN)

MSmits: pretty sure he went the marchete route, but I admit there's a chance I am wrong

MSmits: I think darkhorse is the first non-NN on that list

MSmits: possibly Chibikko but i doubt it

MSmits: Look at it this way: re curses bot was 5 rating points ahead of any non-NN for months. I only got near him by counterbooking him. Without that I was 3-4 rating below him and 1-2 above everyone else

MSmits: (except you and robo ofc)

MSmits: I PM-ed renard so I guess we'll know

Default avatar.png Risodu: Hi Automaton2000

Default avatar.png Automaton2000: what you think about it

MSmits: no lewd suggestions Automaton2000 or we will recode you

Default avatar.png Automaton2000: that's why i needed to do some research

MSmits: don't research it!

Default avatar.png MalaysiaGamer_Fish: hello

Thorcode: hi Automaton2000

Default avatar.png Automaton2000: but the trick is to use a neural net

Default avatar.png Mayankch: pikachu

jacek: :upside_down:

struct: hi

ScienceGuy66: :) :( :) :(

ScienceGuy66: :nerd::nerd::nerd::nerd::nerd:

Chibikko: > possibly Chibikko but i doubt it NN, AlphaZero-like without any tricks

jacek: what are tricks for alphazero? :o

Default avatar.png PringleOrange: hi, is it possible to lose CP for getting a bad score in something?

struct: yes

Chibikko: I don't know. Maybe endgame book?

jacek: i see

jacek: so normal alphazero like

MSmits: so top 15, everyone is nn except darkhorse and me?

MSmits: I think even some below top 15 are nn :P

jacek: only cool kids use nn

MSmits: apparently :)

struct: !ban NNs

Chibikko: No. NNs are the future

MSmits: still feels a bit like outsourcing your own brain

MSmits: no need to even understand the game you're coding a bot for, except to minimal depth

MSmits: there's something sad about that

jacek: still you need design the nn

MSmits: yes there's new interesting parts about this, but it's a different skill

MSmits: you lose something real when moving from old-style bots to nn's

MSmits: and you gain something else

actualCoderTrevor: Anyone know if the Spring Challenge will go back to the old format?

Default avatar.png PringleOrange: Other than line racing and mad pod racing, what's an easy bot programming thing that I can do for easy CP?

struct: Line racing maybe

jacek: he said that ~

struct: ah lol

jacek: reading mod :unamused:

struct: maybe hypersonic

struct: or you can just make a mcts and the use it on all board games

Astrobytes: hi struct, how's the sfml/imgui going?

struct: I paused it for a bit, after I tried a few things on the eval

Astrobytes: any improvement?

struct: my bot had like 50% win rate with no eval vs my current one for some reason

Astrobytes: lol

jacek: huh

struct: or it was 50% by only evaling pawns

struct: or something

struct: maybe the way I do the eval on ept is wrong

Astrobytes: not sure there is a wrong way per se, as long as you're consistent

jacek: you can be wrongly consistent

Astrobytes: wrong consistently or consistently wrong? :P

struct: after I reach certain depth on the rollout I do

struct: eval(p0) - eval(p1)

struct: or the inverse based on who owns the node

Astrobytes: score range?

struct: 0-1

struct: for each eval

struct: so it can be -1, 1

struct: But its always very close to 0 I think

struct: thats the problem I have

jacek: eval(p0) != -eval(p1) ?

struct: wdym jacek?

Astrobytes: zero-sum eval

Astrobytes: minimax-style

jacek: whats in eval(..)

jacek: eval(p0) and eval(p1) should be related somehow no?

struct: so I should try to score based on if p0>p1

struct: they are the same

Astrobytes: treat it just like minimax eval

struct: but p0 evals p0 and p1 evals p 1

jacek: ok i think its alright

Astrobytes: usually with EPT I return something other than -1,0,1 tbh, don't think it matters much though

Scarfield: i dont think it matters, you have to adjust the exploration constant though

Astrobytes: yep

Scarfield: oi

Astrobytes: hehe how's things?

Wontonimo: :wave: Scarfield !

Scarfield: good good how bout you? :)

Scarfield: hey wonto

Astrobytes: same ol' crap, different day, can't complain :)

**Astrobytes throws a taco in the air

Ninilin: hey dont throw the food

Scarfield: Astrobytes cat is happy for food on the floor

Astrobytes: It's for Wontonimo to catch. Taco Golf

Scarfield: xD

Wontonimo: xD

Wontonimo: :large_taco_accepted_icon:

Astrobytes: :D

Default avatar.png b4d-53ct0r: hellooo

jacek: tacocat

Default avatar.png b4d-53ct0r: morocco

Default avatar.png b4d-53ct0r: any one here from morocco ?

Default avatar.png Siham_ich: Im algerien but my mom she's from marocco

Default avatar.png b4d-53ct0r: pm

Default avatar.png Marshal_1: hello

Default avatar.png Marshal_1: new here

ProCrazyCoder: old here

Default avatar.png Marshal_1: aww never old

ProCrazyCoder: yeah

Default avatar.png devnatiofra: Sus

Husoski: ProCrazyCoder Old enough to remember punched cards? (I am!)

ProCrazyCoder: Husoski :) I'm old enough to remember floppy disk

Husoski: Close...Floppies were invented about the time I started coding.

ProCrazyCoder: Hah, I would like to look at punched cards

jacek: why would you punch cards? what did they do?

ProCrazyCoder: interesting question

ProCrazyCoder: Maybe they are first SSD)

Husoski: Here's a look at GitHub, circa 1970: https://www.ibm.com/ibm/history/ibm100/images/icp/C854721S34852J03/us__en_us__ibm100__punched_card__hand_cards__940x727.jpg

Uljahn: i remember those were called perforated cards (perfocards)

Husoski: Not in the US

Uljahn: i mean in USSR

ProCrazyCoder: i think USSR have punch cards after US

Husoski: Here, they were often called "IBM cards" because IBM made most of the equipment. (...and *invented* most of the techology.)

Husoski: Even DEC/Univac/etc. shops used IBM keypunch equipment for data entry.

Husoski: Mainframe makers made their own card readers and punched card output devices, though...

Dren: I feel like associating a thing with the brand/company that makes it is a uniquely US practice though

ProCrazyCoder: Xerox:)

Husoski: Google?

Husoski: Hand me a kleenex, I'm gonna sneeze!

ProCrazyCoder: oh yeah i know what is kleenex

ProCrazyCoder: but in russia we do not have it

ProCrazyCoder: What do you think about 30nm russian processor Elbrus?

Husoski: Don't know about it. Is is pin compatible with anything that has an affordable motherboard?

Uljahn: better than nothing :slight_smile:

ProCrazyCoder: https://filearchive.cnews.ru/img/news/2020/05/06/mb605.jpg

jacek: 30nm? what year is it

ProCrazyCoder: 2021:)

Husoski: Wikipedia's en-us article on Ebrus-8S say it's a 28 nm processor. Other sources say the Ebrus-8C is either 28 or 24 nm.

Husoski: Too small for me to see, even using my "reading in bed" glassis!

ProCrazyCoder: Yes it 28nm but it's very far from AMD or Intel

Blokops: i hate homework so much im procastianitng by doing practice problems

ProCrazyCoder: But anyway i think its important many others thing like CPU cache and so on

ProCrazyCoder: Blokops lv22 seems to me that you are not procrastinator

Blokops: I can do programing problems without any issues but when i need to do a 10 page essay is when the problem start

ProCrazyCoder: language essay?

Blokops: programing theory

Blokops: need to find a paper about machine lerning

Blokops: and make an essay about it

ProCrazyCoder: that's seems interesting to me even it's not easy

ProCrazyCoder: anyway everything will be paid later

ProCrazyCoder: I ve got a project in the university about ML

Andrisssd: hey guys, is it easy to switch c# to php?

ProCrazyCoder: switch in life?

Andrisssd: like start learning php with c# knowledge

ProCrazyCoder: for my experience i ve been learning c++ and used a little php, and it seemed to me that there is the same logic in loops, conditions and so on

ProCrazyCoder: but for more deeper difference in them better ask more experienced coders

Andrisssd: waiting for some answers from more experienced coders already :) but thank you for attention

Default avatar.png ToasterKeyboard: Hello! Is there a way to increase the difficulty when problemsolving?

KiwiTae: ToasterKeyboard yes you can do it while drunk or stuf like that

KiwiTae: o/

KiwiTae: or try to solve harder problems

Default avatar.png ToasterKeyboard: Is it possible to skip the easier problems in the problem trees? I found that most are around CF Div2 A and B

5DN1L: What problem trees? Anyway, you can always choose whatever puzzle you want to solve here:

5DN1L: https://www.codingame.com/training

5DN1L: Or compete with otheres in any of these:

5DN1L: https://www.codingame.com/multiplayer

5DN1L: ... of these: bot programming, code golf, optimisation

jacek: maybe i should give abalone a chance :thinking:

Default avatar.png m_onako: Anyone up for some help about basic task?

Default avatar.png m_onako: Unary

Dren: unary?

Default avatar.png m_onako: Chuck norris task with 0, it says unary

Dren: what do you need help with?

Default avatar.png djaiswadi: hi

Dren: hello

Default avatar.png Dehumanized: i figured it out, didnt read the task well enough :P

Dren: now try to make it in one line of code

Dren: lol

Dren: ah wc

Thorcode: Hi Automaton2000

Default avatar.png Automaton2000: meh: last weeks d&b has two newcomers in top3 and interestingly now everyone else is in a loop

Thorcode: Automaton2000 who code you?

Default avatar.png Automaton2000: yes, you need to keep it simple

Thorcode: do you know how to reach legend in CBS Automaton2000

Default avatar.png Automaton2000: if you take into account the fact that i don't use it anymore

Blokops: Automaton2000 who is your favorite user in this chat

Default avatar.png Automaton2000: so i figured it out

Blokops: Automaton2000 well tell us if you figure it out

Default avatar.png Automaton2000: only you can answer that

Blokops: :flushed:

Default avatar.png ibrahim123: hi

Default avatar.png ibrahim123: abcdefgHIJKlmonpwsxz\

Default avatar.png ibrahim123: 1111111

Default avatar.png ibrahim123: 1

Default avatar.png ibrahim123: 1

Thorcode: Automaton2000 who is your favorite user in this chat

Default avatar.png Automaton2000: especially if you are looking at the output

Thorcode: Automaton2000 who is your favorite user in this chat

Default avatar.png Automaton2000: hey dont throw the food

Thorcode: Automaton2000 who is your favorite user in this chat

Default avatar.png Automaton2000: i'm curious how much you can do it

Default avatar.png Narujo: 44444444/////////////////////////////////////////////////////////

Default avatar.png Narujo: .01

C26_1: Hmmm, I want to make a binary calcultor

C26_1: calculator*

C26_1: where we have 4 basics operator

C26_1: ^ & ~ |

C26_1: That represents for XOR, AND, NOT and OR operators of 2 nputs

C26_1: inputs*

C26_1: and each inputs is decimal

Blokops: what are you using

Blokops: langauge

C26_1: @Blokops, Python

C26_1: And I'm gonna search how can we made calculators using code

C26_1: And after that.....

C26_1: Wait

Blokops: just do bin(number)[2:]

Blokops: no wait better

C26_1: I mean the input looks like this

Blokops: result = num1 & num 2 print(bin(result)[2:]

Blokops: that will work

C26_1: ~5 = 010 = 2

C26_1: @Blokops, ok

Blokops: no need to bianry before operation

Blokops: & | and ^ are aready in python

Blokops: is like +, -, *

C26_1: And parethesis, I forgot

C26_1: (5^2)&(~3|5)

C26_1: It should look like this if you want

C26_1: Wait for me a second

Blokops: you can abuse the eval function

Blokops: basically you can pass a string and eval will take it as a legal operation

Blokops: eval("1024 + 1024")

Blokops: like so