EPTThe Big Programming Thread - Page 989
| Forum Index > General Forum |
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. | ||
|
WarSame
Canada1950 Posts
| ||
|
Excludos
Norway8263 Posts
On January 14 2019 08:31 solidbebe wrote: ? Graduates from technological universities have an average starting salary around median income here. I'm sorry, bad choice of words from me there. It's far from useless, but it's going to be on the low end for the industry. My first job only 4-5 years ago gave me about half of what I earn now, and I started out a bit higher than most of my friends. It's still true that it was well above median income tho. | ||
|
Acrofales
Spain18328 Posts
On January 14 2019 03:23 travis wrote: I want to execute a function every 5 minutes - within a time window of a few seconds. It needs to be consistent even if it runs for 24 hours straight. It needs to match the time on a website. I can sync my windows clock to the time on the website. Does anyone know how to do this in python? edit: im gonna use APScheduler https://apscheduler.readthedocs.io/en/latest/
| ||
|
Manit0u
Poland17774 Posts
Don't you just love meaningful and non-confusing naming? | ||
|
Deleted User 3420
24492 Posts
On January 14 2019 20:50 Acrofales wrote:
my concern there is that it would lose accuracy over time. like, the first time if i ran it exactly at 12 am, by 24 hours later it might be off by like 5 seconds? or would that not happen | ||
|
Acrofales
Spain18328 Posts
On January 14 2019 22:50 travis wrote: my concern there is that it would lose accuracy over time. like, the first time if i ran it exactly at 12 am, by 24 hours later it might be off by like 5 seconds? or would that not happen You'd have to profile it, but my guess is that launching a thread is negligible (in the range of 1 millisecond or less). In 24 hours you launch 24*12 = 288 threads, so if it's 1 ms for launching a thread, your inaccuracy is still well under a second. That said, if you already do it with a scheduler, then don't worry about it. | ||
|
Deleted User 3420
24492 Posts
(which will likely make it readily apparent I know nothing) when it comes to like, reinforcement learning - models are based on a dataset, yes? and the model gets updated over and over and over, hence the "reinforcement"? - feel free to cut me off if this is way wrong but if that is correct so far, then my question is how do these datasets not grow so large that they are impossible to use? like, alphago for example? am I vastly underestimating how clever programmers are about this stuff? is the magic in how neural nets work? | ||
|
Lmui
Canada6223 Posts
On January 15 2019 10:15 travis wrote: here's a question (which will likely make it readily apparent I know nothing) when it comes to like, reinforcement learning - models are based on a dataset, yes? and the model gets updated over and over and over, hence the "reinforcement"? - feel free to cut me off if this is way wrong but if that is correct so far, then my question is how do these datasets not grow so large that they are impossible to use? like, alphago for example? am I vastly underestimating how clever programmers are about this stuff? is the magic in how neural nets work? Training the neural net requires computation power equivalent to the size of the dataset but calculation of an output based on the input takes a fixed amount of processing power. Pretty much O(n) for training, O(1) for calculation | ||
|
Deleted User 3420
24492 Posts
but I want to throw out one last question about neural nets this is purely a guess - are neural nets basically graphs where nodes are possible actions (or choices... or classifications... or whatever) and edges determine choice sequences, and the weights of edges determine the odds of following choice 1 into choice 2? edit: no, no it isn't. not at all! but my idea could be something neat | ||
|
aRyuujin
United States5049 Posts
| ||
|
Acrofales
Spain18328 Posts
On January 15 2019 11:01 travis wrote: okay, ill stop asking questions before I annoy people because I obviously should just be taking an actual course on this but I want to throw out one last question about neural nets this is purely a guess - are neural nets basically graphs where nodes are possible actions (or choices... or classifications... or whatever) and edges determine choice sequences, and the weights of edges determine the odds of following choice 1 into choice 2? edit: no, no it isn't. not at all! but my idea could be something neat Your idea sounds kinda like a Markov Decision Process (MDP). Neural networks (ANN) are something different. Perhaps you're confused, because you can use an ANN to solve an MDP, and you saw something like that? Anyway, to answer your earlier question: RL generally doesn't learn from a dataset, but from trial and error. You could use a dataset as the basis for the reward function, but the idea is really that it executes some policy (method for determining the best actions), gets some outcome and uses this to update the policy, then uses the new policy to do it all again. This is a time-consuming process, but as Lmui pointed out, once you have finished training, and have a good policy, execution of the policy is straightforward. | ||
|
LightTemplar
Ireland481 Posts
Also want to agree with the folks above. Focusing on a specific language / stack has always seemed like a pitfall to me. All the best people I've worked with will pick up the tools for the job when it makes sense to do so; some of the most frustrating are those that refuse to try something new. | ||
|
ZenithM
France15952 Posts
And like everything else, there is a spectrum of skill and expertise within AI/ML, after a while it won't be worth much to put that on your CV if all you did was take a class. I would say at this very moment it indeed looks good on your CV though. Even small businesses are looking for a dedicated "data scientist" even if they don't realize they need one full-time guy to do that. | ||
|
Manit0u
Poland17774 Posts
This is also why there's currently such a big market for people in those fields. Companies are hiring whomever had any contact with it, regardless of their actual knowledge on the matter or experience with doing anything that's running live in production and making money. I have a friend who's working for a company that wants to release a new product. They are forcing introduction of blockchain into the project despite every developer telling them that they don't need blockchain for their use case scenarios. It's really sad and I hope that this stupid hype bubble will burst soon so that people will finally realize that you can't use ML and blockchain for everything... | ||
|
Excludos
Norway8263 Posts
edit: Also this http://joelgrus.com/2016/05/23/fizz-buzz-in-tensorflow/ | ||
|
ZenithM
France15952 Posts
 . | ||
|
LightTemplar
Ireland481 Posts
Glad the blockchain into everything and AI on everything isn't just my experience. Sort of.... | ||
|
waffelz
Germany711 Posts
On January 16 2019 02:53 LightTemplar wrote: Amazing :D Glad the blockchain into everything and AI on everything isn't just my experience. Sort of.... Have you some data on this? A lot of data preferably... like a big pile of data? I just realised, ML is the logical conclusion for all the big data hacks (not to discard big data as a whole ofc). Think about it, we got lots of cases where we couldnt make sense out of big data so now we throw ML at it so it can make no sense out of it. I expect to get abducted any seco | ||
|
Deleted User 3420
24492 Posts
What is a good way to do that? Brain is pretty tired. I am using python but I am sure I could convert non python code to python if you want to write it in your favorite language (without relying on libraries). | ||
|
Excludos
Norway8263 Posts
On January 16 2019 07:11 travis wrote: Given a 2d point, and a distance, I would like to output a point every 10 degrees on the circle that surrounds that point by the distance. It can start at 0 degrees, I don't really care. So, 35 points. What is a good way to do that? Brain is pretty tired. I am using python but I am sure I could convert non python code to python if you want to write it in your favorite language (without relying on libraries). This seems more like a mathematical problem than a programming one. Once you find the formula for moving around the circumference of a circle (easily found by googling if you don't want to have a go at it yourself), actually implementing the solution is a cakewalk. Edit: Or, download, install and learn Unity (and C#), and use the function transform.RotateAround()  | ||
| ||
