Can You Solve This Without Trial and Error?

For fans of Kakuro

Re: Can You Solve This Without Trial and Error?

Postby denis_berthier » Mon Dec 16, 2013 10:23 pm

saul wrote:
denis_berthier wrote:Could any one of you try M44252 ?

I did it without much difficulty:
r2c2 - r4c4 + r5c5 - r7c7 = 8
Bill pointed out this cut to me in a previous post about a puzzle with this same configuration of cells, if I'm not mistaken. Anyway, it was a configuration very like this, with the the four cells in the running along the main diagonal.

Thanks. This again suggests that the atk rating is based on linear programming.

Mani, are you using anything similar in your solution?
denis_berthier
2010 Supporter
 
Posts: 1253
Joined: 19 June 2007
Location: Paris

Re: Can You Solve This Without Trial and Error?

Postby qwerty13 » Mon Dec 16, 2013 10:42 pm

Is it possible to make your solvers SudoRules and KakuRules public ?
Thank you in advance
qwerty13
 
Posts: 2
Joined: 16 December 2013

Re: Can You Solve This Without Trial and Error?

Postby denis_berthier » Tue Dec 17, 2013 3:10 am

qwerty13 wrote:Is it possible to make your solvers SudoRules and KakuRules public ?
Thank you in advance

I've had this question many times - and my answer has always been the same: no, sorry.
I'm using these solvers (and their common CSP-Rules core) as a research tool. Making them public would require at least:
- developing some simple user interface,
- writing a user manual.

Moreover, these solvers implement my own pattern-based view of solving without T&E (with my specific models and sets of rules). As you're new on this forum (AFAIK), I'm not sure this is what you are looking for.
But I think you can find many online solvers on the web.
denis_berthier
2010 Supporter
 
Posts: 1253
Joined: 19 June 2007
Location: Paris

Re: Can You Solve This Without Trial and Error?

Postby Smythe Dakota » Tue Dec 17, 2013 4:40 am

denis_berthier wrote:Could any one of you try M44252 ?
In response, saul wrote:I did it without much difficulty:
r2c2 - r4c4 + r5c5 - r7c7 = 8

Bill pointed out this cut to me in a previous post about a puzzle with this same configuration of cells, if I'm not mistaken. Anyway, it was a configuration very like this, with the four cells in the running along the main diagonal.

I'd bet dollars to donuts (actually, that's only about an even bet, isn't it?) that the two puzzles are isomorphic.

Bill Smythe
Smythe Dakota
 
Posts: 534
Joined: 11 February 2006

Re: Can You Solve This Without Trial and Error?

Postby Mani » Tue Dec 17, 2013 5:48 am

Denis,

"Thanks. This again suggests that the atk rating is based on linear programming.

Mani, are you using anything similar in your solution?"

No, I don't use any formal solving methods. I haven't got them listed down too, they are ingrained, instinctively, in the mind. Often, the challenge is to avoid arithmetic errors, especially when doing surface computations.
Mani
 
Posts: 33
Joined: 13 December 2013

Re: Can You Solve This Without Trial and Error?

Postby saul » Tue Dec 17, 2013 2:41 pm

denis_berthier wrote:Thanks. This again suggests that the atk rating is based on linear programming.

Sorry, I don't follow. What is it that suggests this to you?
saul
 
Posts: 105
Joined: 01 February 2013
Location: Kansas City

Re: Can You Solve This Without Trial and Error?

Postby denis_berthier » Wed Dec 18, 2013 2:29 am

saul wrote:
denis_berthier wrote:Thanks. This again suggests that the atk rating is based on linear programming.

Sorry, I don't follow. What is it that suggests this to you?

The equation you use: r2c2 - r4c4 + r5c5 - r7c7 = 8
Using only combinations of digits allowed by the sectors, it can't be solved without T&E.
denis_berthier
2010 Supporter
 
Posts: 1253
Joined: 19 June 2007
Location: Paris

Re: Can You Solve This Without Trial and Error?

Postby saul » Wed Dec 18, 2013 1:50 pm

Smythe Dakota wrote:I'd bet dollars to donuts (actually, that's only about an even bet, isn't it?) that the two puzzles are isomorphic.

You win a doughnut, but there's no way to post it. Actually, the isomorphism is the identity. We discussed this puzzle at some length in a series of posts starting June 26. That's when you pointed out the surface sum.
saul
 
Posts: 105
Joined: 01 February 2013
Location: Kansas City

Re: Can You Solve This Without Trial and Error?

Postby Mani » Thu Dec 19, 2013 5:21 am

To Saul's trial-and-error theme, I like to add puzzle H62352 of ATK. The puzzle lends itself easily to solving the greater part and leaves a tough upper right quadrant resisting closure. I resorted then to a clear and limited trial-error strategy choosing a 4 in one of two possible cells in a 45 vertical starting at R8C1, to reach an impossible solution. Choosing then the 4 in the other cell helped cracked the sum. Perhaps it is possible to arrive at this solution without trial-error but the extent of visualisation of resulting cell values goes so deep as to challenge reasonable human competence.

In such cases, I am quite happy to launch a clear and limited trial-error strategy, without choosing a purist approach all the way!

I can provide a snapshot of the position I reached before this if someone can advise how to copy paste the snapshot into the discussion forum.
Mani
 
Posts: 33
Joined: 13 December 2013

Re: Can You Solve This Without Trial and Error?

Postby saul » Thu Dec 19, 2013 3:05 pm

Mani wrote:I can provide a snapshot of the position I reached before this if someone can advise how to copy paste the snapshot into the discussion forum.


The way I've done it in the past is to post the picture on dropbox, as I explained in an earlier post. However,while answering your
question, I found a better way.

Below the text entry area, there is a tab labeled "Upload attachment." Click this and choose the file you want to upload. Then click on "Add the file." A button titled "Place inline" will appear. Click this also. You can repeat the process if you want to add more than one picture. You will see the file names in your post, surrounded by HTML starting and ending "attachment" tags. It seems like this will be inserted at the point of the insertion cursor in the text entry, but you can cut and paste this text to have the attachment appear elsewhere in the text.
saul
 
Posts: 105
Joined: 01 February 2013
Location: Kansas City

Re: Can You Solve This Without Trial and Error?

Postby denis_berthier » Thu Dec 19, 2013 9:57 pm

Mani wrote:To Saul's trial-and-error theme, I like to add puzzle H62352 of ATK.
[...]
Perhaps it is possible to arrive at this solution without trial-error but the extent of visualisation of resulting cell values goes so deep as to challenge reasonable human competence.

A tough one indeed, but still solvable without T&E (it is in W9).

I'm not sure to understand your last remark. As this puzzle requires whips[9], any T&E solution will have at least one elimination with a larger step, i.e. with a sequence of at least 2x9 inferences (9 negative + 9 positive).
If you mean that, for such puzzles, no one cares for finding the "simplest" solution (i.e. using the smallest chains), I agree.
denis_berthier
2010 Supporter
 
Posts: 1253
Joined: 19 June 2007
Location: Paris

Re: Can You Solve This Without Trial and Error?

Postby saul » Thu Dec 19, 2013 11:34 pm

denis_berthier wrote:I'm not sure to understand your last remark.

I think what Mani is saying is that it might be possible to do this mentally by building up the inference chain in your head, but it probably contains too much information for most of us to remember.
saul
 
Posts: 105
Joined: 01 February 2013
Location: Kansas City

Re: Can You Solve This Without Trial and Error?

Postby denis_berthier » Fri Dec 20, 2013 1:37 am

saul wrote:
denis_berthier wrote:I'm not sure to understand your last remark.

I think what Mani is saying is that it might be possible to do this mentally by building up the inference chain in your head, but it probably contains too much information for most of us to remember.

Sure, as the chain length increases, it becomes more and more difficult to keep it only in mind. One needs to write links on the board.
The main difference I see between a whip (or braid) with target some candidate C and T&E based on C is, in T&E one doesn't have to remember how other candidates are asserted or eliminated when propagating the consequences of C; but I doubt that T&E for C can be done mentally (correct me if I'm wrong - I've never tried).
denis_berthier
2010 Supporter
 
Posts: 1253
Joined: 19 June 2007
Location: Paris

Re: Can You Solve This Without Trial and Error?

Postby Mani » Fri Dec 20, 2013 4:23 am

Oh yeah, I did mean that it is too much to remember and visualise a chain of cell values that may result from one assumption, it would be easier to save a position and try out a T&E sequence to save time and avoidable brain-crunching, limiting such T&E where one is clear that 2 or 3 alternative assumptions is all that will take to reach the right solution.

No one has replied to my request, how to paste kakuro grids as Saul has done before in this trail.
Mani
 
Posts: 33
Joined: 13 December 2013

Re: Can You Solve This Without Trial and Error?

Postby Mani » Fri Dec 20, 2013 5:18 am

I would add H31452, which I just solved, where again such T&E or simulation was called for.
Mani
 
Posts: 33
Joined: 13 December 2013

PreviousNext

Return to Kakuro