@EvilJS huh? which post? did you save the url?

think you had some viable ideas on the matrix-vector multiplication that seemed to work but your writing has difficulties with clarity (and, full sympathy, but plz dont use foreign language as excuse here). you cant assume anyone knows anything on stackexchange, its part of the point to be specific. hyperlinking wikipedia in your answers is std/ helpful.

@vzn yeah, no excuses. And that answer that you are reffering to is waiting to be finally improved. I am still working on algorithm, so final edit should contain algorithm and code.

@vzn but to be honest you are assuming that talking to me in my native language is somehow easier. No it is definitely not ;)

@vzn deleted answer

@vzn I had written why he should not compare it, and where difference came from. He was freshly registered to downvote was not an option. I do not downvote and do not want to participate in witch hunt or get any kind of revenge, just wanted to know what was wrong with answer to prevent it from reccuring. But there was no comment, so I deleted it.

If anybody is feeling that my post need edit, or should be deleted. I give my blessing and my help with that.

@EvilJS lol actually was not assuming that & was suspecting otherwise, as am hinting above :P

@vzn algorithm for now builds up pairs of bits (per column, something like [1, 1, 0, 0]) then derives triplets and generates subtraction of one column, gives +1 to cost, and so on. After deriving up to row size, there is built something like CYK but for all possible permutations.

and why did you delete it?

And then starts the fun, like in CYK I am feeding rows to get derivation, but taking counters with penalties as objective function to find minimum.

you deleted it because of single downvote? what are you, a girl? :P

@vzn well, assumed (my bad) that it was due to other plans like closing it.

@vzn when you ask like this, you are not first to state this question.

@vzn let me think.

is that your 1st downvote dude? oh! have an idea. you cant see total up/ downvotes so maybe you never had a negative overall vote so far? because think you have downvotes on other posts but maybe not visible in the total score (sum of up/ downvotes)

@ no. Girls cry a lot I don't ;)

did you answer the question with downvotes? its at -2 currently. thats kinda half your problem right there :P

@vzn not first one, I got one on 4D line and did not deleted answer. Just haven't figured out better algo yet.

ok. life goes on. looking fwd to upvoting more of your contributions to the site :)

@vzn yes, it is downvoted, I answered when it was 0. When counters started to break below -1 I deleted it. Initial thought - question like this are ontopic, abstract and full of misconceptions, let us save the world ;)

@vzn well, upvote is +10, downvote -2, easy to spot on parity check and small intervals. Also I find out a way to check who did it, but it's a waste of time ;)

Good old hacking tricks come handy. Still using one on gmail ;)

@vzn by "glides" on your blog in topic Collatz, it means that from some number there are several jumps?

Still haunted ;)

Cursing is not allowed here?

@EvilJS a glide refers to iterates that "stay above" (are greater than) the starting "seed". if one can prove all trajectories are finite-length "glides" the conjecture is proved. glides are a initial sequence of the total trajectory & the remaining iterates in the trajectory are below the starting seed. (thx for perusing the blog, alas do not get much feedback on it.)

What should I write then? Hi, it is me, good blog, cheers, or that I tried several times to do something in the topic and never thought of gliding distance, as it fluctates to much to clear out my ideas ;)

Never got blog, so I do not know what it's like. I have written some, but never used them after release.

@vzn have you seen it: yet another blog

@EvilJS yes, a neat page with lots of empirical study, once cited roosendaal here after reading about him in lagarias' writeup/ survey. the combination of the two gave me some ideas to study glides in some ways. actually did not even refer to them as "glides" before that page probably. & have long wished lagarias or roosendaal would join SE :)

some blogs have heavy attn & comments, others do not, it can take years to build up a readership. but sometimes even elite scientists do not get much attn or comments. and nonscientists can have very highly read blogs etc.... its mass "cyber" populism in action...

@vzn Vox populi, vox Dei ;)

roosendaal may have done some of the most thorough empirical experiments on collatz around, but he seems to be cited "not that much".

I found once paper with very clear intro: Knowledge forgotten. It was about Sylvester resultants. It's usage is near to nonexistent. And people using delta for quadratic equation do not know where it came from.

@vzn ever tried something with polynomial exact root finding? By exact I mean also symbolic and checking if two roots are identical (even if not representef by radicals). I am asking you (and here), because it is strictly connected to CS (and further implementation), and I would really like trying something implementable ;)

I have been fighting with Risch algorithm for long time. I have implemented probably more than there is in Axiom (open source now, Computer Algebra System), and honestly took several ideas from it's source. But I hit the wall, without good polynomial solver it is perfectly useless.

@EvilJS not following. risch algorithm is for integration. root finding can only be done numerically for higher-degree polynomials. how does one do "exact symbolic polynomial root finding"?

@vzn yes Risch algorithm is to integrating. Polynomial root finding is subproblem. That is the point, I do not know how it should be done, but there are some neat transforms like Tschirnhaus, Bring, some forms of two variable functions to represent root.

By exact e.g. x^5 -x +1, let me name it's root v5, I do not have numerical value, but when I encounter the same value, I would like to make it comparable. It is like sqrt(2) or Pi, store them as a string.

So I am interested in any kind of transform or series (infinite) and encode it as string, than compare it with another one in terms of this transform. There are for sure some awesome transforms or theorems that I do not know. So searching without luck. And approximation is something that I do not want to do.

I know Galois theorem, and ideas how to represent polynomials up to degree 10. As far this is still finite, some decent idea to encode comparable roots in any notation up to 27 degree inclusive are sufficient for me. And any general idea like Viete even not practical, but implementable is what I am looking for. For now, I do not know magic keywords, so even name to follow in that direction and ask proper questions is a booster.

@vzn changing subject, you said once that you were doing some psychology, could you elaborate what exactly?

@EvilJS think it would help a lot if you just describe the applied problem & then we work backwards from that. or are you just experimenting with math techniques?

did not say was "doing psychology" said have studied it.

symbolic (vs typical numeric) root finding seems to me rarely attempted. maybe some techniques implemented in mathematica or other symbolic math pkgs? maybe more a question for Mathematics? or Computational Science?

@vzn ok, maybe I shall agree to all of the above, but simple question what have you studied is rather not worth changing chat ;)

@vzn with Risch - implementing for fun, exploring possibilities, when algorithm fails in computing integrals that are elementary, but there is poor polynomial solver, and error is there, I am exploring

Backtracking to problem is not an option, Axiom documentation is dozen thousands pages long, and code I have is not faulty, it simply misses good polynomial system

Prefer Axiom over other packages as it is well overdocumented with codes, and I thought here is good place, as I am interested in implementamble algorithms, that even are not fully working, not theorems that it is impossible ;)

Maybe I am overconfident in thinking that CS solves it, but let me take BSplines, it is defined that 1/0 = 1. Nobody cries and it works :)

@vzn ok. Went there and found awesome topics. Maybe my writting is unclear, but reading is good. BBL.