Anyone here do anything in machine learning/neural nets/Deep Learning?
Is it worth looking into, or is it just more memes?
>>57479066
I do. What do you want to know about it?
>>57479066
Basically I see two paths in ML/Deep Learning right now. They aren't well-defined but rather a loose hunch of what I see.
ML Engineers: Use ML algorithms, big data technologies to solve business relevant problems. There are tons of books, blogs and resources to get started. Minimum requirements are programming abilities in Python, Java, R, etc. or an ability to learn. Additionally basic knowledge in Statistics.
ML Researcher: Pushing the bounds of ML knowledge. Produces things industry has yet to adopt or is actively adopting. These are academics. PhD students, or PhD holders. Require solid foundation in mathematics/statistics and specialization in a ML field.
I'm a high school student working on creating a neural network to solve the rubiks cube. I know it's a long shot but do you think it is practically impossible since I'm going to have 40 thousand billion billion different possible inputs? (all the different legal cube configurations)
>>57480833
im no expert but i dont think thats how it works.
Why dont you just have 9 x 6 inputs. one for each square.
>>57480833
i think it greatly depends on how you use the net for. NN are generally best at recognizing or estimating things, so unless you pair it up with an euristical search i don't think you'll gonna do much.
instead of bothering training a neural net you could just calculate the 3d manhattan distance, wich is long to code, but actually O(n) fast.
>>57480833
This won't work, just use a standard algorithm.
i thought 3D manhattan distance was just |x2 - x1| + |y2 - y1| + |z2 - z1|
It's what I do for work... It gives you the ability to let the computer solve difficult problems. I think it is useful in nearly all fields of computing. Www.udacity.com have some good free courses for beginners
>>57483159
>It gives you the ability to let the computer solve difficult problems
I like machine learning
but the amount of buzzwords etc are cringeworthy
>>57479066
does creating an AI-waifu count?
>>57483599
rip tay
may kek have mercy on her soul
I'm working in big data at the moment, a client asked our company to adapt a log management system we wrote with machine learning.
I don't know almost anything about ML, don't know if i should pick up this job, it seems interesting tho. With a good knowledge of scala should i be able to learn quickly?
>>57479074
How'd you learn it?