Need advanced method to solve

Post the puzzle or solving technique that's causing you trouble and someone will help

Re: Need advanced method to solve

Postby SpAce » Fri May 15, 2020 10:18 pm

denis_berthier wrote:There can't be any fixed order. CSP-Variables come first, then their value, so that you will always have some rn{c... and some cn{r... and some bn{xx.. So, the order rn or nr is irrelevant.

That conclusion doesn't follow from the premises, nor is it true.

As someone else (maybe Red Ed) said of it someday, it is the inventor's prerogative (and he did mean me).

Of course. You're totally free to use whatever notation you like. I only suggested one very simple way that could make it more accessible to Eureka users (and probably most other people too). You don't have to agree or even care.

Each one is a target in turn. (They are never used simultaneously in the chain.)

Sounds really inefficient but that's a feature of typical contradiction eliminations anyway, which is one reason why they're inelegant. I hope your biv-chains don't have the same limitation? If they do, the only reasons for that I could see are a) it's a limitation of SudoRules and you've made a virtue out of necessity, or worse b) you actually treat biv-chains as contradiction chains on the logical level too (which would mean they're not really equivalent to AICs but to the obsolete Discontinuous Nice Loops).

Who is "we" and "ours"?

The regular Eureka+extensions users on this forum. That would be the majority, with maybe 1-2 individuals for each of Nice Loops, implication chains, TDP, your notation, and some individualistic ones. Of course I'm mostly talking about the regular players in the Puzzles section, as it's kind of hard to know what notations others possibly use. The same trend is evident in the historical posts too. Nice Loops with their horrible notation seem to be the standard in the oldest posts until succeeded by AICs and Eureka. There's very little evidence of anything else having had much foothold here, even though your notation pops up every now and then (mostly posted by you).

There was no "memory chains" before I introduced whips, braids and all my menagerie.

Didn't I just suggest that possibility? Like I said, I don't know anything about that history so I can't know if you're right or not. If you are, good for you. All I know is that me, myself, and I would never knowingly take credit for anyone else's ideas, and if I do know the original inventor of some technique (almost impossible in many cases) I will mention it when applicable. You should know that even based on this discussion alone.

That said, I'm not responsible at all if at some point in history some people have picked up some good parts from your system without giving due credit to you, or if such credits were lost in the sands of time. I feel for you, if that's true, but there's nothing I can do about it. I have no way of knowing who invented what, having been on the scene for only three years, and so many of the old discussions having been lost in the crash. It's hardly relevant to my quest of finding the best solving and notation tools anyway.

Some people pretend not to understand them and propose "new" chains" and new names for them or for related concepts as if they had invented anything.

Well, I don't know who those people are, or if they exist, so I really couldn't care less. This discussion is getting less and less interesting.

There is no "our" system. There is just an alternative notation that tries to hide the total absence of new ideas or plagiarism.

Ok. Since it seems to be your primary concern, let's just agree that you've probably invented every useful sudoku technique especially related to chains. Everyone else has just plagiarized you and tried to hide that fact under alternative (and of course totally illegible) notations. It's obviously impossible that different people might have come up with similar ideas independently, even if those ideas seem pretty inevitable extensions to their existing mechanisms (or did you invent AICs too, I forgot to ask?). It's also totally impossible that they might have implemented them with more readable and intuitive notations.

Would that agreement stop this broken record that's probably been playing for years for different audiences? Seems like the simplest solution to me, which goes well with the "simplest first" theme of this thread.

One last thing, though. Even if it's true that the concepts in memory chains were stolen from you, it's pretty irrelevant in the big picture, perhaps even something you shouldn't be too proud of. Most of us try to avoid them as much as possible as they're considered very inelegant. It's actually always possible, but sometimes, though very rarely (and for some people never), a memory chain might seem like the optimal compromise. Thus, fact is that at least the best Eureka writers use almost exclusively techniques and notations that probably have nothing at all to do with your inventions.

For example, I haven't seen anything resembling things like split-nodes or nested chains in your system. With those we can write almost anything as perfectly valid AICs. No ugly contradictions or memories needed at all, thank you very much. (Are they always the shortest or the most readable chains? Probably not, but that's not the point. Is recursion the most understandable and efficient way to program?) As far as I understand, you don't have things like krakens either because you don't accept OR-branching. Many times those provide the most readable solutions compared to any linear chains implementing the same logic. So, if you really want to claim memory chains as your invention, be my guest :)

I'm not interested at all in talking about notation. It's pure waste of time.

Let's be done with wasting each other's time then. I was genuinely interested in learning your system but I no longer see much point. Thanks anyway, and I mean it. I do appreciate even your minimal willingness to help. I now have a much better idea of your system, and at least I learned what whips are at last (or did I just pretend?).
User avatar
SpAce
 
Posts: 2514
Joined: 22 May 2017

Re: Need advanced method to solve

Postby denis_berthier » Sat May 16, 2020 12:07 am

SpAce wrote:
denis_berthier wrote:I'm not interested at all in talking about notation. It's pure waste of time.

Let's be done with wasting each other's time then. I was genuinely interested in learning your system but I no longer see much point.

That seems to be the right conclusion. Your only interest was notation, which is a pure waste of time if it conveys no new idea.
denis_berthier
2010 Supporter
 
Posts: 1620
Joined: 19 June 2007
Location: Paris

Re: Need advanced method to solve

Postby SpAce » Sat May 16, 2020 7:55 am

denis_berthier wrote:Your only interest was notation

Another false conclusion. It's true that I'm more interested in notational details than most, but it's ridiculous to conclude that it's my only interest. My primary interest is to learn to solve as efficiently as possible by being able to pick the best parts from as many competing paradigms as possible. I just happen to realize that without a proper understanding of the relevant languages it's impossible to learn and to communicate ideas effectively and efficiently.

Thus, to understand your system, or any system, I must first learn its language. Once I do, it's usually a short step to understanding everything else behind it. In case you haven't noticed, when I put my focus on something I'm usually pretty good at seeing the essence of it, even behind carefully crafted smoke screens or otherwise unnecessary complexities. Sometimes that essence is not quite as spectacular or unique as it's been claimed or thought to be. Of course, pointing that out doesn't make me very popular amongst the prophets and the true believers.

which is a pure waste of time if it conveys no new idea.

That's false too, unless it includes ideas relating to the notation itself. It's the same as claiming that it's a waste of time to study and develop more intuitive or efficient programming languages, or mathematical notations and any other domain specific languages, or even to study human languages. Sure, all of those are secondary to the ideas they're supposed to communicate, but they do have intrinsic value and measurable properties which either improve or hinder their primary function.

Besides, it seems to me that notation is only irrelevant to you when someone doesn't agree that yours is the best or even dares to point out possible improvements for it. Otherwise you seem very eager to assert (not as a subjective opinion but as a fact) how other systems (like TDP) lack proper notation (I mostly agree, btw) or how AICs or Allan Barker's sets are totally illegible (I don't agree, though both could be improved). So, you should probably make up your mind whether notation matters to you or not, instead of flip-flopping depending on whether it serves your agenda or not.
User avatar
SpAce
 
Posts: 2514
Joined: 22 May 2017

Re: Need advanced method to solve

Postby denis_berthier » Sat May 16, 2020 8:26 am

Hi, SpAce,
You should have stuck to your previous conclusion.
As for me, I'll do.
Harassing each of us in turn won't lead you very far.
denis_berthier
2010 Supporter
 
Posts: 1620
Joined: 19 June 2007
Location: Paris

Re: Need advanced method to solve

Postby SpAce » Sat May 16, 2020 10:15 am

denis_berthier wrote:Hi, SpAce,
You should have stuck to your previous conclusion.
As for me, I'll do.
Harassing each of us in turn won't lead you very far.

Or you. Otherwise you're absolutely right, Denis. I'm just not very good at letting falsehoods fly, especially if they include assumptions about me and my motives. In this case I just as well could have, though, because it was so obvious and not very serious anyway. No need to "harass" you any further on my part.
User avatar
SpAce
 
Posts: 2514
Joined: 22 May 2017

Re: Need advanced method to solve

Postby eleven » Sat May 16, 2020 12:45 pm

Just a word about "inventing" memory chains.
The most primitve solving method using a basic solver is to set one of the candidates in a cell, look, if it causes a single somewhere, set the single and so on. What is produced that way ? A memory chain.
eleven
 
Posts: 2461
Joined: 10 February 2008

Re: Need advanced method to solve

Postby denis_berthier » Sat May 16, 2020 1:52 pm

eleven wrote:Just a word about "inventing" memory chains.
The most primitve solving method using a basic solver is to set one of the candidates in a cell, look, if it causes a single somewhere, set the single and so on. What is produced that way ? A memory chain.

No. What is produced that way is a step of T&E.
When I introduced the t-chains and zt-chains on the forums of that time, there was a deluge of hatred because I had broken their overarching rule of having only reversible chains.
denis_berthier
2010 Supporter
 
Posts: 1620
Joined: 19 June 2007
Location: Paris

Re: Need advanced method to solve

Postby eleven » Sat May 16, 2020 2:16 pm

denis_berthier wrote:
eleven wrote:Just a word about "inventing" memory chains.
The most primitve solving method using a basic solver is to set one of the candidates in a cell, look, if it causes a single somewhere, set the single and so on. What is produced that way ? A memory chain.

No. What is produced that way is a step of T&E.

Yes, and if this step of T&E leads to an elimination, it is called a memory chain (in fact it is also a memory chain without elimination, but it is just useless).
Last edited by eleven on Sat May 16, 2020 2:38 pm, edited 1 time in total.
eleven
 
Posts: 2461
Joined: 10 February 2008

Re: Need advanced method to solve

Postby denis_berthier » Sat May 16, 2020 2:34 pm

eleven wrote:
denis_berthier wrote:
eleven wrote:Just a word about "inventing" memory chains.
The most primitve solving method using a basic solver is to set one of the candidates in a cell, look, if it causes a single somewhere, set the single and so on. What is produced that way ? A memory chain.

No. What is produced that way is a step of T&E.

Yes, and if this step of T&E leads to an elimination, it is a memory chain.

Where is the memory in eliminating a candidate ?
denis_berthier
2010 Supporter
 
Posts: 1620
Joined: 19 June 2007
Location: Paris

Re: Need advanced method to solve

Postby eleven » Sat May 16, 2020 2:40 pm

The memory is reminding all the singles in this primitive chain (allowing more singles later). This does the simple solver for you.
I really don't know any manual solver, who did not "invent" memory chains on his own. The problem is, that it becomes harder and harder to remember all the earlier steps, when the chain becomes longer. That might be the reason, that i never saw a manual solver post solutions in your notation (they always came from people, who had implemented the SudoRules).
eleven
 
Posts: 2461
Joined: 10 February 2008

Re: Need advanced method to solve

Postby denis_berthier » Sat May 16, 2020 2:52 pm

eleven wrote:The memory is reminding all the singles in this primitive chain (allowing more singles later). This does the simple solver for you.
I really don't know any manual solver, who did not "invent" memory chains on his own. The problem is, that it becomes harder and harder to remember all the earlier steps, when the chain becomes longer. That might be the reason, that i never saw a manual solver post solutions in your notation (they always came from people, who had implemented the SudoRules).

Except that what you're describing is T&E, the most widely used method.
It takes a lot of work to transform this into different types of chains with various continuity conditions, and still more to find the shortest ones.

But if you're happy in your world of imprecise ideas, that's fine for me. Everything you like is an AIC. Everything you don't like is T&E. Everything else is a memory chain.
denis_berthier
2010 Supporter
 
Posts: 1620
Joined: 19 June 2007
Location: Paris

Re: Need advanced method to solve

Postby eleven » Sat May 16, 2020 3:20 pm

denis_berthier wrote:Except that what you're describing is T&E, the most widely used method.
It takes a lot of work to transform this into different types of chains with various continuity conditions, and still more to find the shortest ones.

But if you're happy in your world of imprecise ideas, that's fine for me. Everything you like is an AIC. Everything you don't like is T&E. Everything else is a memory chain.

False, false and false.
Yes, your chains are based on T&E, and your program finds and optimizes them. But i just stated, that you cannot claim to have invented memory chains.
I am the last one, who only likes AIC's.
And the last one, who says, that everything else is a memory chain.

And now i let you have the last word.
eleven
 
Posts: 2461
Joined: 10 February 2008

Re: Need advanced method to solve

Postby denis_berthier » Sat May 16, 2020 3:50 pm

eleven wrote:
denis_berthier wrote:Except that what you're describing is T&E, the most widely used method.
It takes a lot of work to transform this into different types of chains with various continuity conditions, and still more to find the shortest ones.

But if you're happy in your world of imprecise ideas, that's fine for me. Everything you like is an AIC. Everything you don't like is T&E. Everything else is a memory chain.

False, false and false.
Yes, your chains are based on T&E, and your program finds and optimizes them. But i just stated, that you cannot claim to have invented memory chains.
I am the last one, who only likes AIC's.
And the last one, who says, that everything else is a memory chain.
And now i let you have the last word.


Thanks. I'll take this opportunity for the last word. And be it really the last one if you have nothing intelligent to say.

I haven't invented "memory chains" which are not even defined. Only the name is used to refer to ideas first introduced in my chains.

In terms of resolution rules, what I invented is whips, braids, g-whips, g-braids, their typed (i.e. 2D in Sudoku) counterparts...
And the difference with memory chains, which remain devoid of any precise content, is, my chains give rise to theorems, to statistical analyses, to a precise relation (which is very far from identity) between braids and T&E. Can you refer to anything of this kind for "memory chains"?

My chains are not based on T&E. The only "program" in SudoRules is the rules themselves, as they are expressed in PBCS (syntax slightly different). They are not optimised from T&E, they are directly found by pattern-matching, a radically different process.
denis_berthier
2010 Supporter
 
Posts: 1620
Joined: 19 June 2007
Location: Paris

Re: Need advanced method to solve

Postby SpAce » Sat May 16, 2020 4:26 pm

This is not a reply to anyone. Just to lighten up the mood, this Schrödinger's cat is an extreme example of what can be done with memory chains. No matter who invented them, I'm pretty sure that neither that person nor anyone else envisioned such crazy possibilities where a candidate's ghost memory keeps operating next to its dead body :D Steve's comment nailed it:

SteveG48 wrote:That's pretty weird, all right, using 3 as being both true and not true in the same term. You're collapsing the wave function from undefined to both dead and alive at the same time. Even Schrödinger might call it a day.
User avatar
SpAce
 
Posts: 2514
Joined: 22 May 2017

Previous

Return to Help with puzzles and solving techniques