/sci/ - Science & Math » Thread #10874225

797KiB, 1280x1280, 1280px-Lorenz_attractor_yb.svg.png

View Same Google iqdb SauceNAO

Anonymous Thu 08 Aug 20:07:15 2019 No.10874225 View Reply Original Report

Quoted By: >>10874239 >>10874243 >>10874252 >>10874504 >>10874538 >>10874631 >>10875819 >>10877659

How do I into chaos theory?

I want to see if i can analyze neural networks using chaos theory stuff, maybe find latent variables somehow by training maybe. But I may be pseuding myself into some corner here cuz i can't into chaos theory. Hilfe!

Anonymous

Anonymous Thu 08 Aug 2019 20:15:53 No.10874239 Report

Quoted By:

>>10874225

>How do I into chaos theory?

learn about differential equations

Anonymous

View Same Google iqdb SauceNAO antitechrevolution.jpg, 2MiB, 1125x1500

Anonymous Thu 08 Aug 2019 20:19:41 No.10874243 Report

Quoted By: >>10874259 >>10874276

>>10874225
Read pic related

>Even if we make the wildly improbable assumption that the behavior of our society could be predicted through the manipulation of, say, a million trillion simultaneous equations and that sufficient computing power to conduct such manipulation were available, collection of the data necessary for insertion of the appropriate numbers into the equations would be impracticable, especially since the data would have to meet impossibly high standards of precision if the predictions were expected to remain valid over any considerable interval of time. Edward Lorenz, a meteorologist, was the first to call widespread attention to the fact that even the most minute inaccuracy in the data provided can totally invalidate a prediction about the behavior of a complex system. This fact came to be called the "butterfly effect" because in 1972, at a meeting of the American Association for the Advancement of Science, Lorenz gave a talk that he titled "Predictability: Does the Flap of a Butterfly's Wings in Brazil Set Off a Tornado in Texas?" Lorenz's work is said to have been the inspiration for the development of what is called "chaos theory" -the butterfly effect being an example of "chaotic" behavior.

Anonymous

View Same Google iqdb SauceNAO received_1676838052459697.jpg, 19KiB, 400x403

Anonymous Thu 08 Aug 2019 20:21:52 No.10874252 Report

Quoted By: >>10874266 >>10874271

>>10874225
Why chaos theory? You know they have actual textbooks over neural network modelling, right? Plus, a lot of the contents of hidden layers aren't really interpretable by humans, so what do you mean by "analyze"?

Anonymous

Anonymous Thu 08 Aug 2019 20:26:19 No.10874259 Report

Quoted By: >>10874262 >>10875889 >>10875891

>>10874243
>never heard of Cambridge Analytica
They used Big Data to influence voters bud, these people can definitely predict you. That's what Facebook and Google use your data for.
>being against AI
So, you want a future with dog shit technology and dog shit medicine? Contain it to yourself. Some of us actually want a decent future.

Anonymous

View Same Google iqdb SauceNAO TED I TRIED TO WARN YOU.png, 158KiB, 406x395

Anonymous Thu 08 Aug 2019 20:28:48 No.10874262 Report

Quoted By: >>10874464 >>10875880 >>10876926

>>10874259
Technology won't give humanity a decent future, retard.

>But when all people have become useless, self-prop systems will find no advantage in taking care of anyone. The techies themselves insist that machines will soon surpass humans in intelligence. When that happens, people will be superfluous and natural selection will favor systems that eliminate them-if not abruptly, then in a series of stages so that the risk of rebellion will be minimized.

>Even though the technological world-system still needs large numbers of people for the present, there are now more superfluous humans than there have been in the past because technology has replaced people in many jobs and is making inroads even into occupations formerly thought to require human intelligence. Consequently, under the pressure of economic competition, the world's dominant self-prop systems are already allowing a certain degree of callousness to creep into their treatment of superfluous individuals. In the United States and Europe, pensions and other benefits for retired, disabled, unemployed, and other unproductive persons are being substantially reduced; at least in the U. S., poverty is increasing; and these facts may well indicate the general trend of the future, though there will doubtless be ups and downs.

Anonymous

View Same Google iqdb SauceNAO mathematics is not worthwhile.png, 1MiB, 1239x1758

Anonymous Thu 08 Aug 2019 20:30:47 No.10874264 Report

Quoted By:

Anonymous

Anonymous Thu 08 Aug 2019 20:31:29 No.10874266 Report

Quoted By: >>10874537 >>10874645

>>10874252
>You know they have actual textbooks over neural network modelling
Yes, but does that cover _why_ neural networks work? Most of the stuff I've studied so far has been how to create solutions which work, but very little has been about why it actually works. And further experiments I've done has shown that the architecture of the models doesn't matter at all as much as the amount of parameters which are being trained.

>Why chaos theory?
Because I'm always amazed how looking at the patterns of a cup of tea which i just poured milk into is basically completely unpredictable. Simple things which becomes insanely complex are interesting. If i can understand/grasp that, or at least understand and predict that (maybe using neural networks/similar machine learning), I might be able to predict all sorts of things. Also: in order to create better neural networks.

Anonymous

Anonymous Thu 08 Aug 2019 20:33:36 No.10874271 Report

Quoted By: >>10874309 >>10874645 >>10874654

>>10874252
>Plus, a lot of the contents of hidden layers aren't really interpretable by humans
Missed this. This is my point though. The latent factors should be interpretable with sufficient understanding in chaos theory.

My theory is that neural networks work because we create our semi-controlled chaos system which's properties we train. But, seeing that I don't know anything about chaos theory, I don't know if this is the case.

Anonymous

Anonymous Thu 08 Aug 2019 20:39:29 No.10874276 Report

Quoted By:

>>10874243
lets say we make 7 billions human clone and move out of the solar system. wed have a running simulation of earth at a sensible economic price and computing power.

Anonymous

Anonymous Thu 08 Aug 2019 20:59:29 No.10874309 Report

Quoted By: >>10874315

>>10874271

So you know nothing about chaos theory, yet you think it can explain how a neural network works.

You've never opened a serious statistics or math textbook, have you?

Neural networks aren't some kind of black magic that nobody understands.

If you want to work on something interesting, then develop a theory that rigorously explains how genetic algorithms work instead.

Anonymous

Anonymous Thu 08 Aug 2019 21:02:42 No.10874315 Report

Quoted By: >>10874318

>>10874309
>You've never opened a serious statistics or math textbook, have you?
Nope, that's what I'm asking for.

>Neural networks aren't some kind of black magic that nobody understands.
Where can I find out how they work?

Anonymous

Anonymous Thu 08 Aug 2019 21:05:05 No.10874318 Report

Quoted By: >>10874321

>>10874315

Nonlinear Dynamics and Chaos by Steven H. Strogatz

Pattern recognition and machine learning by Christopher M. Bishop

Anonymous

View Same Google iqdb SauceNAO 1543418409962.webm, 643KiB, 720x720

Anonymous Thu 08 Aug 2019 21:09:57 No.10874321 Report

Quoted By:

>>10874318
Thank you, have a webm.

Anonymous

Anonymous Thu 08 Aug 2019 22:41:12 No.10874464 Report

Quoted By:

>>10874262
So, you must be posting on 4chan from your Amish farm then?

Anonymous

View Same Google iqdb SauceNAO TIMESAND___LISSAJOUS.gif, 301KiB, 201x201

Anonymous Thu 08 Aug 2019 23:06:09 No.10874504 Report

Quoted By:

>>10874225
>How do I into chaos theory?
Write a script to make a animated plot of the Lorenz attractor. Then mess around with the initial condition a little see what it does. Then read about the Lyapunov exponent. After that, you know about a quarter to a half of all of chaos theory.

Anonymous

Anonymous Thu 08 Aug 2019 23:19:07 No.10874537 Report

Quoted By:

>>10874266
That's not really Chaos Theory, anon. You seem to be more interested in Statistics and possibly Probability. For neural networks, you can study Bayesian Statistics and Mathematical Probability or even look into Statistical Learning. Also, understanding "why" they work is diving into what computing is in terms of what the brain does exactly (which is a cutting-edge research field), so you're getting deep into Artificial Intelligence and possibly neuroscience.

Anonymous

Anonymous Thu 08 Aug 2019 23:19:23 No.10874538 Report

Quoted By:

>>10874225
if you know calculus and some linear algebra, pick up strogatz "nonlinear dynamics and chaos"

Anonymous

Anonymous Thu 08 Aug 2019 23:50:41 No.10874631 Report

Quoted By:

>>10874225
>I want to see if i can analyze neural networks using chaos theory stuff
nice bait

Anonymous

Anonymous Thu 08 Aug 2019 23:58:16 No.10874645 Report

Quoted By:

>>10874266
>>10874271
you have no idea how retard you sound, hurry up and pick book to understand why are you embarrassing yourself

Anonymous

Anonymous Fri 09 Aug 2019 00:04:02 No.10874654 Report

Quoted By: >>10875833

>>10874271
>My theory is that neural networks work because we create our semi-controlled chaos system which's properties we train.
What you're really interested in then is the line/relationship between chaotic and stochastic systems. There is a relationship of sorts between """chaos theory""" and """machine learning""", but it's a fairly trite one.

Anonymous

Anonymous Fri 09 Aug 2019 14:54:22 No.10875819 Report

Quoted By: >>10875836

>>10874225
it just has a nice sounding name it ins't really all that interesting.

Literally just sensitivity to initial conditions. same initial condition = same result , slightly different initial condition = very different result.

Anonymous

View Same Google iqdb SauceNAO sha256.jpg, 165KiB, 857x749

Anonymous Fri 09 Aug 2019 15:03:10 No.10875833 Report

Quoted By: >>10875855

>>10874654
Is sha-256 a chaotic or stochastic system? In the eyes of mathematicians, I mean. If no: is sha-512? Where does the line go?

Anonymous

Anonymous Fri 09 Aug 2019 15:04:27 No.10875836 Report

Quoted By:

>>10875819
>same initial condition = same result ,
What makes it interesting to me is that there's infinitely many slightly different conditions. What really matters?

Anonymous

Anonymous Fri 09 Aug 2019 15:21:57 No.10875855 Report

Quoted By:

>>10875833
>Is sha-256 a chaotic or stochastic system?
It's a highly chaotic function, yes.

Anonymous

Anonymous Fri 09 Aug 2019 15:45:44 No.10875880 Report

Quoted By:

>>10874262
>In the United States and Europe, pensions and other benefits for retired, disabled, unemployed, and other unproductive persons are being substantially reduced
excellent, a return to the natural order

Anonymous

Anonymous Fri 09 Aug 2019 15:57:46 No.10875889 Report

Quoted By: >>10876512 >>10877660

>>10874259
People are way more complex than a computer can predict. Neurologists have yet to even understand the human brain fully.
Furthermore, even the US government has tried brainwashing people using LSD under MKUltra and did not succeed. My point is that humans are not machines that can be told what to do entirely, though they can be suggested on what to do.

Anonymous

Anonymous Fri 09 Aug 2019 15:58:49 No.10875891 Report

Quoted By: >>10876906

>>10874259
Influencing the people who actually bother to vote is about as impressive to influencing brute animals.

Anonymous

Anonymous Fri 09 Aug 2019 22:43:59 No.10876512 Report

Quoted By:

>>10875889
neurologists dont study how the brain works, they study the causes of brain illness and how to treat it.

Anonymous

Anonymous Sat 10 Aug 2019 01:34:33 No.10876906 Report

Quoted By:

>>10875891
Cambridge Analytica's trick was influencing people who would not normally bother to vote.

Anonymous

Anonymous Sat 10 Aug 2019 01:43:54 No.10876926 Report

Quoted By:

>>10874262
>When that happens, people will be superfluous and natural selection will favor systems that eliminate them-if not abruptly, then in a series of stages so that the risk of rebellion will be minimized.

What a prescience.

Anonymous

Anonymous Sat 10 Aug 2019 10:45:57 No.10877659 Report

Quoted By:

>>10874225
https://colah.github.io/

Anonymous

Anonymous Sat 10 Aug 2019 10:45:57 No.10877660 Report

Quoted By:

>>10875889
You don't need to understand them fully, you just need the person's data relevant to your problem. If the profiling methods you are using for the problem at hand are doing poorly, the learning model will adjust itself and try new approaches until it finds a good profiling method. That is the magic of data science with social engineering and psychology.

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