💾 Archived View for dioskouroi.xyz › thread › 29398147 captured on 2021-11-30 at 20:18:30. Gemini links have been rewritten to link to archived content
-=-=-=-=-=-=-
________________________________________________________________________________
People have posted (math/compsci) proofs on 4chan .... Post it to the internet. Lol
https://www.wired.com/story/how-an-anonymous-4chan-post-help...
Go through the hoops to put a preprint on Arxiv. Put in the effort to properly format the document, etc. They’ll probably want you to get an endorser before you can submit, but they also explain how to find people to endorse you.
Are you certain you have not rediscovered a previously known algorithm? a common variant of Dijkstra's algorithm finds the shortest path from one node to all other nodes in a graph, using in O((V + E)log v) time and O(V + E) space (and can actually get a better big O with a Fibonacci heap, but no one uses this in practice).
So from reading your post, I'm not entirely sure what sets your algorithm apart!
If you _do_ want to make claims about your algorithm being state of the art, I recommend trying to come up with a formal proof for the run time, and type up a manuscript in latex presenting the algorithm, a proof of correctness / runtime, and placing the algorithm into the existing literature. You can then try to send that to academics to read over, or submit to a conference as a solo author, if the work is good I don't think they will ignore you.
>Find the shortest path from one node to all other nodes in a graph
That is precisely what Dijkstra's algorithm does. The limitation of that algo is that it only finds one path from the start to each other node, not all possible paths. As a really simple example, consider a 2x2 graph:
A---B
| |
C---D
If you want to go from A to D, Dijkstra will return _either_ A-B-D or A-C-D, but not both. The "k path problem" tries to find all paths from A to D
Ah, I see what you are saying now. Have you seen this
https://stackoverflow.com/questions/2819347/dijkstras-algori...
stack overflow post yet? It seems to be provide a way to modify Dijkstra's to do what you want?
I'm not sure this is what you are looking for, but if that variant of Dijkstra's can be made to work you should probably include it as a comparison in any white paper you make!
It looks like you're not afraid of publicly talking about your discovery. For that reason, I recommend going to the papers you read with previous solutions and send everyone an email explaining what you discovered (the ones who are alive). Academics are very aproachable if they know what they're doing (AKA not famous).
If they respond to you, ask them if they would be willing to review your draft. Write a draft of a whitepaper in the format of abstract, content, conclusion and send it to them attached. Only if they agree to review to you, if they claim that it's their discovery you have paper trail to prove otherwise and they'll lose their reputation, they won't do that. Try to get feedback of as many people you can, trying to prove your discovery have some flaws or that it was already discovered by somebody else.
If you have enough evidence that it's a new discover, now next step is to ask them for contacts of some credible comp-sci publication that you can publish it. Add them as thank you for reviewing your draft and publish it. After you publish, post in HN with the name of the journal for credibility.
If you want to spend some money and go hard on defending the thing as "yours" before having definitive evidence from other scholars that it's really novel, register a patent. It's similar standard to that of the standard of an academic white-paper published in a journal.
I don't recommend, as patents are today useless for innovation and serious work, although that's the exact reasons they were created in the first place
Thanks! This all makes sense to me
The devil is usually in the details, are you sure you're not missing a 'k' factor in your O() complexity?
In your shoes I'd try to set up some benchmarks and other algorithms and compare, just for the learning experience. Try and fiddle with large V/E/k factors and compare. Check out some profiling tools, etc.
If all goes well you could just publish a library, that should count as prior work, nobody can come and steal your glory if you do wish to make it public..
Im almost certain I am not missing any k factor. I've been carefully checking for that. The algorithm would have fallen apart on any unweighted graph with hundreds of nodes with at least 2 connections each.
EDIT: PS I read the paper you linked before your ninja edit :)
Can't you just produce an article with LaTeX and self-publish the article in relevant forums, a la the Bitcoin whitepaper? Maybe find some researches with published articles about this problem and send it to them asking for comments. If it is indeed a breakthrough it will make the rounds.
Or do you specifically want to publish in some journal?