/sci/ - Science & Math » Thread #14263066

13KiB, 304x263, imagedrawtrendline1.jpg

View Same Google iqdb SauceNAO

Would you qualify linear regression as a form of machine learning?

Anonymous Fri 25 Feb 17:41:16 2022 No.14263066 View Reply Original Report

Quoted By: >>14263167 >>14263567 >>14265858 >>14266318 >>14266650 >>14268314 >>14269837

Is drawing a trend line really considered a kind of machine learning? Because anyone with excel can do as such. Does that mean the bar for machine learning is really low or are computers are just that advanced?

Anonymous

View Same Google iqdb SauceNAO 798bb08de689f47a05d83a3101a706ea.jpg, 93KiB, 850x1192

Anonymous Fri 25 Feb 2022 17:45:41 No.14263073 Report

Quoted By: >>14263077 >>14263229 >>14267844

>is (generic algorithm) machine learning???
no

Anonymous

Anonymous Fri 25 Feb 2022 17:48:49 No.14263077 Report

Quoted By: >>14263089 >>14263238 >>14263970 >>14265819

>>14263073
By that logic, no algorithm is machine learning, which obviously isn't true. If I implement Bayes Algorithm and predicted the stock market using that, that would be machine learning dumbass.

Anonymous

Anonymous Fri 25 Feb 2022 17:55:20 No.14263089 Report

Quoted By:

>>14263077
>By that logic, no algorithm is machine learning, which obviously isn't true
does not follow, retard
>If I implement Bayes Algorithm and predicted the stock market using that, that would be machine learning dumbass.
lol

Anonymous

Anonymous Fri 25 Feb 2022 18:41:14 No.14263167 Report

Quoted By: >>14266651

>>14263066
Yeah. Machine learning is a meme term. All statistical modeling is machine learning by definition. Linear regression is the fundamental machine learning problem, everything else is just the nonlinear techniques.

The problem of machine learning is literally solving Y = F(X) where F is unknown. What's simpler than starting with the linear case to build your methods? The linear case has an exact analytical solution, making it computationally cheap as fuck and is why many techniques will even use linear regression at some point within their overall methods.

Anonymous

Anonymous Fri 25 Feb 2022 19:11:32 No.14263200 Report

Quoted By:

it's the most basic form of analysis, but the principles behind it are basically the same as some ridiculous deep net. ML (or """""AI""""") is just abusing massive amounts of data in order to achieve predictive power, that's it. Is linear regression the same as RGN? No, but it's the same concept.

Anonymous

Anonymous Fri 25 Feb 2022 19:18:47 No.14263205 Report

Quoted By:

Linear regression is ML, more widely known as AI. So you can create AI with pen and paper, isn't computer science cool?

Anonymous

Anonymous Fri 25 Feb 2022 19:29:03 No.14263229 Report

Quoted By: >>14263247

>>14263073
in what way is linear regression a genetic algorithm

Anonymous

Anonymous Fri 25 Feb 2022 19:38:31 No.14263238 Report

Quoted By: >>14263451

>>14263077
Bayes would be better suited for something like sentiment analysis in a situation like this. For example: if you have a chat room or social media discussion about a particular stock, you can make a model to give each post/discussion about a stock a sentiment (positive, negative, neutral) where you can then add up all of the sentiments to give an overall sentiment (positive, negative or neutral outlook) for said stock.

Anonymous

Anonymous Fri 25 Feb 2022 19:57:09 No.14263247 Report

Quoted By:

>>14263229
in what way is the word genetic generic?

Anonymous

Anonymous Fri 25 Feb 2022 22:05:22 No.14263451 Report

Quoted By:

>>14263238
Thanks.

Anonymous

View Same Google iqdb SauceNAO le mao.png, 38KiB, 1234x1302

Anonymous Fri 25 Feb 2022 23:01:40 No.14263567 Report

Quoted By: >>14263595 >>14263965

>>14263066
>Would you qualify linear regression as a form of machine learning?
yes

Anonymous

Anonymous Fri 25 Feb 2022 23:17:28 No.14263595 Report

Quoted By: >>14265800

>>14263567
What is that glossary?

Anonymous

Anonymous Sat 26 Feb 2022 01:38:19 No.14263802 Report

Quoted By:

The bar isn't that low, but not really far off. The more you see these things the more you realize their limitations. We don't have the theory for strong AI. I think an AI like Mike from The Moon is a Harsh Mistress is most similar to how a strong AI might actually come about in the near future.

Anonymous

Anonymous Sat 26 Feb 2022 03:59:15 No.14263965 Report

Quoted By:

>>14263567
Needs a line for statistics: variance/machine learning: <blank>

Anonymous

Anonymous Sat 26 Feb 2022 04:03:02 No.14263970 Report

Quoted By:

>>14263077
>lemme apply this peepee poopoo formula in the stock market and make millions
wow daring today

Anonymous

Anonymous Sat 26 Feb 2022 23:25:29 No.14265800 Report

Quoted By:

>>14263595
It's telling you all the crossover terms between statistics and ML. It's honestly the messiest shit I have seen. How there has not been a push for standardized terminology across the fields is beyond me.

Maybe there needs to be an accreditation body or something.

The only texts I've seen that make a good jab at standardizing the terms, are the mixed texts (like statistical learning texts).

Anonymous

Anonymous Sat 26 Feb 2022 23:35:55 No.14265819 Report

Quoted By:

>>14263077
Even your example wouldn't be machine learning. Machine learning is not "computer do complex thing," its a path that one can use to make a computer do something you want it to do

Anonymous

Anonymous Sat 26 Feb 2022 23:57:52 No.14265858 Report

Quoted By: >>14266331

>>14263066
Who gives a shit about this retarded question. Anyways...

I wonder is there research on ways to at least perform linear regression on absurdly high dimensional matrices. Like closed approximation methods and even SGD approximation methods.

Anonymous

Anonymous Sun 27 Feb 2022 04:16:00 No.14266318 Report

Quoted By:

>>14263066
Machine learning is literally just stats dressed up with trendy
>OMG AI!

Anonymous

Anonymous Sun 27 Feb 2022 04:20:01 No.14266331 Report

Quoted By:

>>14265858
No, because it's useless. The more variables there are, the closer you get to multicollinearity between variables which breaks everything down and makes the problem poorly conditioned.

If you have a shit ton of variables do PCA and other dimension reduction techniques.

Anonymous

Anonymous Sun 27 Feb 2022 06:22:13 No.14266508 Report

Quoted By:

it only becomes machine learning when there is a loss function used to train the parameters of the function. linear regression is a dumb example because a line cutting through the middle of a 2 dimensional set of points doesn't really tell you much. you don't need ML to see "hurr durr trend goes up"

Anonymous

Anonymous Sun 27 Feb 2022 09:17:28 No.14266650 Report

Quoted By:

>>14263066
Yes.
>machine learning is a class of algorithms that tackle some task T and increase some performance metric P given experience E, and where P increases on T with more E.

Anonymous

Anonymous Sun 27 Feb 2022 09:18:47 No.14266651 Report

Quoted By: >>14267644

>>14263167
>The linear case has an exact analytical solution, making it computationally cheap as fuck
Matrix inversion isn't fucking cheap. That's why nobody uses the analytical solution, brainlet.

Anonymous

Anonymous Sun 27 Feb 2022 18:18:28 No.14267644 Report

Quoted By: >>14268605

>>14266651
You're right, but the fact that it has an exact analytical solution (a.k.a. it is invertible) means there are a LOT of computational cheats you can use.

The problem has one unique solution, and has NO other minima. You can use fixed point algorithms and quickly get to your solution set.

Linear regression is a well-posed problem and this makes it superior to many many other techniques because many nonlinear problems are not well posed (there are often multiple solutions not to mention multiple local minima). It's why linear regression and all the different methods for linearizing a problem are the first things anyone is taught (or should be taught at least)

Anonymous

Anonymous Sun 27 Feb 2022 19:30:03 No.14267844 Report

Quoted By:

>>14263073
Wrong, linear regression learns information therefore counts as machine learning.

Anonymous

Anonymous Sun 27 Feb 2022 22:19:28 No.14268314 Report

Quoted By:

>>14263066
Yes. It's literally the first thing taught in machine learning courses, it's a fundamental building block of the field.

Anonymous

Anonymous Mon 28 Feb 2022 00:25:04 No.14268605 Report

Quoted By:

>>14267644
True. It's funny to me how engineers instinctively know and do this, but everyone laughs at them for it.

Anonymous

Anonymous Mon 28 Feb 2022 14:45:52 No.14269837 Report

Quoted By:

>>14263066
Sure, you can consider linear regression a form of machine learning if you want.
>Does that mean that knowing linear regression is enough to work with ML.
No, you retard.
Is knowing addition enough to get a job as a mathematician?
Is knowing how to tie your shoes enough to land you a job as a shoemaker?

We really need to include some basic IQ test for this fucking board.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches