I was very interested in creating a minesweeper solver a few years ago. I got it pretty far along, had a tool that was able to solve a complete board. Sometimes it would solve without issue, other...
I was very interested in creating a minesweeper solver a few years ago. I got it pretty far along, had a tool that was able to solve a complete board. Sometimes it would solve without issue, other times it would return probability distributions that didn't make any sense. I wrote a few posts about it on my (now abandoned and horrible looking) blog. My goal was to avoid computing the probability distribution using the current configuration of the board each time a square was revealed. I wanted to use the latest probability distribution and incorporate the "measurement" gathered from the latest revealed square. I think what I was trying to achieved might have been similar to "belief propagation".
Before I could wrap my head around the problems with my implementation, I started a new job and basically abandoned it. I'm still not certain if there was a flaw in my approach or just a bug in my solver.
During that time I did come across somebody's blog with a web-app version of their solver. It's beautiful.
I was very interested in creating a minesweeper solver a few years ago. I got it pretty far along, had a tool that was able to solve a complete board. Sometimes it would solve without issue, other times it would return probability distributions that didn't make any sense. I wrote a few posts about it on my (now abandoned and horrible looking) blog. My goal was to avoid computing the probability distribution using the current configuration of the board each time a square was revealed. I wanted to use the latest probability distribution and incorporate the "measurement" gathered from the latest revealed square. I think what I was trying to achieved might have been similar to "belief propagation".
https://en.wikipedia.org/wiki/Belief_propagation
Before I could wrap my head around the problems with my implementation, I started a new job and basically abandoned it. I'm still not certain if there was a flaw in my approach or just a bug in my solver.
During that time I did come across somebody's blog with a web-app version of their solver. It's beautiful.
https://mrgris.com/projects/minesweepr/demo/player/