|
Thread Rules
1. This is not a "do my homework for me" thread. If you have specific questions, ask, but don't post an assignment or homework problem and expect an exact solution.
2. No recruiting for your cockamamie projects (you won't replace facebook with 3 dudes you found on the internet and $20)
3. If you can't articulate why a language is bad, don't start slinging shit about it. Just remember that nothing is worse than making CSS IE6 compatible.
4. Use [code] tags to format code blocks. |
On March 12 2016 03:17 madnessman wrote:
Has anybody had any success here writing an SC:BW AI which doesn't just execute a few hard coded strategies? I've finished my courses in AI and machine learning and I'm thinking about writing a BW AI to compete in the CIG AI competition this year. I found a few papers where researchers tried to implement more sophisticated methods but those bots didn't fare too well against the other competitive bots.
You could always check out the WC3 AI. It comes bundled with the standard game in the editor. Me and my brother did have a blast editing AI's and pitting them against each other in it. Never got around to checking if you could mess with the internals of it, but it could be worth a try.
|
On March 12 2016 03:25 Manit0u wrote:Show nested quote +On March 12 2016 03:17 madnessman wrote:
Has anybody had any success here writing an SC:BW AI which doesn't just execute a few hard coded strategies? I've finished my courses in AI and machine learning and I'm thinking about writing a BW AI to compete in the CIG AI competition this year. I found a few papers where researchers tried to implement more sophisticated methods but those bots didn't fare too well against the other competitive bots. You could always check out the WC3 AI. It comes bundled with the standard game in the editor. Me and my brother did have a blast editing AI's and pitting them against each other in it. Never got around to checking if you could mess with the internals of it, but it could be worth a try.
There's actually a great FOSS API for interfacing with BW (https://github.com/bwapi/bwapi) which seems to be the standard. Time to brush off the C++ skills!
|
On March 11 2016 22:18 iaretehnoob wrote:
I don't think "strings are character arrays" qualifies as a quirk.
Nah, not that, it's that since php supports heterogeneous arrays you would think an index call to a 1-d array element would not be allowed. Not sure how PHP is implemented but it probably increases overhead having to check to make sure the array element is indexable / that there isn't a runtime error. And yeah as others said it's borderline obfuscatory.
|
C++14 question.
Why do people use auto parameters and auto return type? How is this safe if you change a function in the future?
|
$str = 'abcd';
if (is_array($str)) {
echo 'array';
} else {
echo 'not an array';
} // not an array
echo $str[3]; // d
echo implode(', ', array_values($str)); // ERROR: array_values() expects parameter 1 to be array, string given
![[image loading]](https://i.imgflip.com/10opnk.jpg)
It would seem that strings in PHP are treated like char arrays only when you're trying to access index on them (also used for functions like strpos etc.) but otherwise they're just string literals? It's bullshit if you ask me.
On March 12 2016 06:31 Shield wrote:
C++14 question.
Why do people use auto parameters and auto return type? How is this safe if you change a function in the future?
I guess it's because g++ is a really good compiler and it'll fill those out with proper types anyway? I'm not a C++ specialist by any measure so you'll want a second opinion on this.
|
On March 12 2016 06:36 Manit0u wrote:
$str = 'abcd';
if (is_array($str)) {
echo 'array';
} else {
echo 'not an array';
} // not an array
echo $str[3]; // d
echo implode(', ', array_values($str)); // ERROR: array_values() expects parameter 1 to be array, string given
It would seem that strings in PHP are treated like char arrays only when you're trying to access index on them (also used for functions like strpos etc.) but otherwise they're just string literals? It's bullshit if you ask me. Show nested quote +On March 12 2016 06:31 Shield wrote:
C++14 question.
Why do people use auto parameters and auto return type? How is this safe if you change a function in the future? I guess it's because g++ is a really good compiler and it'll fill those out with proper types anyway? I'm not a C++ specialist by any measure so you'll want a second opinion on this.
C++11 and C++14 have auto defined by standard, so g++ is irrelevant here. Any C++14 compiler is supposed to support this.
|
Yeah, but is it a matter of "Do not help the compiler" (as in - no need to use static typing since the compiler can optimize it nicely for you so you can just go ahead and declare stuff as auto without any loss to performance) or something else entirely?
|
On March 12 2016 03:17 madnessman wrote:
Has anybody had any success here writing an SC:BW AI which doesn't just execute a few hard coded strategies? I've finished my courses in AI and machine learning and I'm thinking about writing a BW AI to compete in the CIG AI competition this year. I found a few papers where researchers tried to implement more sophisticated methods but those bots didn't fare too well against the other competitive bots.
In general, in games, it´s really hard to beat a selection of hardcoded strategies.
Think of chess: sure, you can choose your flavour of opening, but once you choose the opening, you just follow it through for 5-10 moves before things deviate again. In SC:BW this is your build order. Then you get to the hard part of chess, or in SC:BW, you get to the bit where you have to decide based on how the game is developing what to do: army composition, etc. etc. Skirmishes are like chess endgame: there are once again a set of preset moves which work (muta stacking, focus firing, spreading out in an arc, stutter stepping, etc. etc.)
Obviously there are things that are different from chess (pretty much all the decisionmaking is based on incomplete information for starters, and the search space is both far larger and far smaller at the same time).
Anyway, it's a fun project, so go for it! As others mentioned, can do it for WC3 too, which might be more interesting (less figured out because there's so many spells).
|
On March 12 2016 08:47 Manit0u wrote:
Yeah, but is it a matter of "Do not help the compiler" (as in - no need to use static typing since the compiler can optimize it nicely for you so you can just go ahead and declare stuff as auto without any loss to performance) or something else entirely?
I usually refer people to Herb Sutter when talking about auto. The main argument of the linked article is that it abstracts the actual types from your code and makes it dependent only on the interface, which I completely agree with. The thing I think he trivializes a bit is that sometimes it makes your code distinctly less readable when the types become unclear - I tend to use auto everywhere, unless I feel like it would impact readability, for those reasons.
e: maybe an example would be cool.
list<Thing> things = GetThings();
for (const Thing &thing : things) {
thing.DoStuff();
}
auto things = GetThings();
for (const auto &thing : things) {
thing.DoStuff();
}
There are two ways using auto makes this nicer if the return type of GetThings() changes.
1) GetThings() changes to return a vector<Thing> instead of a list<Thing> (because someone you work with heard it was Good Practice (tm) to use vectors and not lists). The explicitly typed code has to be changed to state that things is a vector<thing>.
2) GetThings() changes to return a list<IThing> instead of a list<Thing> (because someone you work with heard it was Good Practice (tm) to use interfaces and not concrete classes). The explicitly typed code has to be changed to state that things is a list<IThing>, and thing is an IThing, while the auto code only relies on the DoStuff() method being present.
|
On March 12 2016 03:17 madnessman wrote:
Has anybody had any success here writing an SC:BW AI which doesn't just execute a few hard coded strategies? I've finished my courses in AI and machine learning and I'm thinking about writing a BW AI to compete in the CIG AI competition this year. I found a few papers where researchers tried to implement more sophisticated methods but those bots didn't fare too well against the other competitive bots.
If you're starting from scratch, you may want to check out UAlbertaBot. It's pretty modular and easy to improve, and has most of the basic software engineering done, things that can be annoying to deal with.
Or you can start completely fresh! As an AI researcher who uses SC:BW on occasion, we'd love to have more people to play against =) As you said, for now good engineering (i.e. limiting manipulatable edge cases) is almost more important than having a sophisticated AI, but hopefully we'll get past that.
|
On March 12 2016 15:18 raNazUra wrote:Show nested quote +On March 12 2016 03:17 madnessman wrote:
Has anybody had any success here writing an SC:BW AI which doesn't just execute a few hard coded strategies? I've finished my courses in AI and machine learning and I'm thinking about writing a BW AI to compete in the CIG AI competition this year. I found a few papers where researchers tried to implement more sophisticated methods but those bots didn't fare too well against the other competitive bots. If you're starting from scratch, you may want to check out UAlbertaBot. It's pretty modular and easy to improve, and has most of the basic software engineering done, things that can be annoying to deal with. Or you can start completely fresh! As an AI researcher who uses SC:BW on occasion, we'd love to have more people to play against =) As you said, for now good engineering (i.e. limiting manipulatable edge cases) is almost more important than having a sophisticated AI, but hopefully we'll get past that.
Thanks for posting this, I'm taking my first AI class next fall and reading through some of the UAlberta code got me super excited
|
Hey guys, my final year design project was a WebGL debugger embedded inside Chrome DevTools.
GitHub repo
Chrome Web Store
If it's not too much trouble, could you guys star the GitHub repo (we can get marks based on stars lol) and if you're super interested in WebGL things, give it an install and a test run? It'd be really helpful.
:D
Also our code isn't super amazing since only one of us is true JS dev, but we think it ended up pretty well.
|
![[image loading]](http://i.imgur.com/LTNUwXK.png)
F**k Zend Framework 2. There is a bug in the code, guess where it is?
No, it's not anywhere in that stacktrace, it's in a controller that doesn't even appear in the stack trace, because of ZF2's stupid event system that detaches itself from the actual code.
If I didn't know what code I touched, I'd be stuck searching the problem for half an hour trying to find the problem.
Don't use ZF2 for your projects.
|
Hyrule18969 Posts
I tried ZF2 a while back and hated it. Laravel and Symfony are sooooo much better, imo
|
On March 16 2016 22:53 tofucake wrote:
I tried ZF2 a while back and hated it. Laravel and Symfony are sooooo much better, imo
Pretty much anything is better. We're still stuck with ZF2, because our dutch overlords used it for the online shop and so we had to use it for our new website as well. Now they are pretty much out of the decision making progress, but we have too much code to scrap it and use something proper.
|
Heh, I'm just negotiating a new contract with the company that wants to turn all of their old ZF1 code into Spring. Might be quite an adventure.
|
On March 16 2016 22:46 Morfildur wrote:F**k Zend Framework 2. There is a bug in the code, guess where it is? No, it's not anywhere in that stacktrace, it's in a controller that doesn't even appear in the stack trace, because of ZF2's stupid event system that detaches itself from the actual code. If I didn't know what code I touched, I'd be stuck searching the problem for half an hour trying to find the problem. Don't use ZF2 for your projects.
All I saw was that your stacktrace starts with pleasure_xxx.php.
Not even hiding that you`re building a porn website? 
|
On March 17 2016 06:13 Acrofales wrote:Show nested quote +On March 16 2016 22:46 Morfildur wrote:F**k Zend Framework 2. There is a bug in the code, guess where it is? No, it's not anywhere in that stacktrace, it's in a controller that doesn't even appear in the stack trace, because of ZF2's stupid event system that detaches itself from the actual code. If I didn't know what code I touched, I'd be stuck searching the problem for half an hour trying to find the problem. Don't use ZF2 for your projects. All I saw was that your stacktrace starts with pleasure_xxx.php. Not even hiding that you`re building a porn website?
It's my job. It pays. And it's not half as fun as you imagine.
|
At least we know it's not pornhub (it runs on Symfony) 
And Morfildur is right, it's just a job. You'd be surprised how soon you get fed up with stuff like that when working on erotic lingerie websites etc. After a while it doesn't really matter what you're displaying, dumb scientology books, gay porn, whatever, you simply stop noticing it.
|
|
|
|
|
|
|
|
EPT
