/sci/ - Science & Math

>>10689246
none, it would just get stolen.

>>10689255
What if I kept the earnings in a bank like, say, Wells Fargo to keep said earnings from being stolen, could it potentially make a lot of money?

>>10689259
not your earnings, your method.

>>10689273
Not if he implements it practically in his own business and keeps it a secret

Like say I sold the algorithm as a "service", where the customers would send the data to the api and they receive the resulting model. Then the algorithm wouldn't get stolen, and I could sell the model as a service.

>>10689286
If you didn't run the algorithm on hardware comparable in power to your competitors then you would need to be getting some sort of insane efficiency gains for it to be competitive with the resources of a larger firm. Foreign agents have been able to steal from Google/Government networks. If you get any sort of attention its gonna get ganked. Your best bet would be to keep it secret until you finish it, get a lawyer, and sell it. Trying to monetize it by yourself probably wouldn't work out well.

>>10689301
That sounds like a good plan. How much money would I likely make if I did this? The method I'm talking about is orders of magnitude more accurate than regression for long term predictions.

>>10689273
What if I kept the method in a bank safe like, say, Wells Fargo on a piece of paper to keep said method from being stolen, could it potentially make a lot of money?

Oh look, it's another episode of anon fucked up his backtesting.

>>10689246

That is already true. Machine learning methods, most of the useful ones at least, are non linear. They already perform better than linear regression for real world data, and no one became rich for inventing those methods.

>>10689510
It isn't messed up backtesting. It exploits properties of functions that regression doesn't.

>>10689275
Method itself notwithstanding, it’s insanely hard to implement an algorithmic trading bot. The computer science required probable constitutes years worth of learning—and forget buying someone else’s implementation—that’s how you get your idea stolen. If OP somehow figured something like this out, I commend him, truly. But he’ll most likely end up being the next Black-Scholes.

>>10689526
*probably

>>10689524
You seem to be saying that linear regression isn't the most effective algorithm in many cases. I don't think you're correct. In the areas where regression is used, a superior method would increase model accuracies

>>10689524
>no one became rich for inventing those methods.
Yeah but what if there was a $1M prize for inventing a method like that? Could you collect a $1M prize for inventing a method like that, in the special case where there is $1M prize advertised?

Say there was a time limit on the prize that said you had to wait 24 months for flaws in your method to be exposed, and it has been 27 months with no flaws exposed and only further confirmations which are like concrete except firmer and more solid. Could you collect the $1M or is getting paid a fuck-you amount of money a meme and they will only let you get money by flicking nickels at you for being their wageslave, or maybe if you grovel to your mom who also is trying to make sure you never get the $1M and got you fired from your last few jobs?

Say you somehow collected the $1M prize, do you think you could use your five minutes of fame to get at least one person to buy your book? What if instead of some stupid machine learning algorithm, you solved something called "the most important unsolved problem in mathematics?" Do you think you could get one or more persons to buy your book?

>27 is a bigger number than 24
>http://www.vixra.org/abs/1703.0073

>>10689525
Nonparametric regression is still regression.

>>10689526
For background, Black and Scholes found an extremely accurate model for options pricing. However, they were unable to implement it profitably in a market that had already found a model that was nearly good enough. So, instead they published their findings and are now among the most well respected academics in finance

>>10689543

Your best bet is joining a machine learning competition (or several), win them since your method apparently is the holy grail and only you have it, and then collect the prizes.
Something tells me you're full of shit, however.

>>10689545
:P

>>10689543
See because actually the reason I ask this is because I disproved the Riemann hypothesis 27 months ago, and there is a $1M prize for that with a 24 month delay. I proved in March of 2017 that there are non-trivial zeros in the neighborhood of infinity. Simply for personal enrichment and some sense of civic duty which drives me to want to contribute to the overall scientific effort driving human progress, I also actually found the zero lat year too. However, the $1M has fucking nothing to do with what numbers are the zeros, the $1M is for proving whether or not such numbers exist. I did that 27 months ago.

Pic is revised for minor edits so date is from 24 months ago.

>>10689589
reddit fags need more sauce... don't

>>10689589
Did you win the prize? Isn't that one of the millenial prize problems?

>>10689599
>>10689600
I didn't win the prize, probably because arXiv rejects every manuscript I send to them. However, it could be because I'm not on the real internet and I'm in a gay NatSec bubble, or it could be that the CIA put code in the PDF so looks legible when I read it but looks different to others. I was wondering if they were doing that int he character encoding on my website too.

Clay Institute extended a waiver to Perelman because he did not publish in a peer-reviewed journal as is required by the rules for the $1M, and for me to get the money for my publication on viXra, then Clay Institute would need to extend to me ~THE~EXACT~SAME~WAIVER~ that they extended to Perelman. However, it seems like everyone would rather pretend like the problem is unsolved than to admit that I'm the one who solved it... because fuck me.

Buy my book though, pic related.

>>10689615
>gay NatSec bubble
The thing about that is, the security of the nation will be just fine as long as I am fine. I can kill all my cousins and their acolytes and the nation will be just fine, but they like to pretend that their lives are important for national security when that just isn't the case.

>>10689616
>gay NatSec

>>10689510
more like anon has nothing to test

>>10689436
Who knows Took? Depends how good it is, and who is buying. Why haven't you driven to a prestigious mathematics department with your papers and asked a respected published professor to look at your work? Or taken it to a large firm with many elite mathematicians and done the same? Why do you NEED to publish through arxiv/vixra?

>>10689546
Only works for euro options though

Zero because it already exists.

Sign your proof using RSA and then post it here along with your public key. If you're as competent as you claim, you will be able to verify your identity using your private key and I'm almost certain 4chan would ensure you're accredited.

>>10690694
I don't want to share the algorithm, I want to make money with it. If I give humanity more powerful AI that will just accelerate their self destruction.

>>10690726
Then make a benevolent AI that leads humanity. That was my plan but if you do it, then you'll save the world.

The AI ought to be rewarded and punished based on the happiness levels of humans.

>>10690734
Perhaps that's my plan anon.

>>10690726
>>10690757
>grandiose delusions
s c h i z o
c h i z o s
h i z o s c
i z o s c h
z o s c h i
o s c h i z

>>10690854
I didn't assert that I had or will make a super AI, so it's not grandiose.

Bump

what's the deal with the schizo posting lately?

>>10689246
It depends how well you could market/implement it.

If you worked for a large finance or tech company the company itself would be able to make trillions.

How much you earn depends on your business acumen, not just your product.