Day 14: Find 2D pattern in matrix #9
Conversation
|
atm the code is not finished, today was a long day and I couldn't bring myself to finish the code. But the thougth process is all there and it's not too complex. |
|
|
||
| there are only 4 possible starting locations: `(0, 0), (0, 1), (1,0), (1,1)`. | ||
|
|
||
| I don't think it's possible to do this problem and faster. So it's not |
There was a problem hiding this comment.
Note: I haven't seen or done this problem before, so this might be totally wrong.
This problem reminds me a lot of fast string matching, which means that there's probably a faster way to do it. In KMP and other linear string search algorithms, a key observation is that you might be able to skip steps. For example,
needle = 'abcd'
haystack = 'abcfffffabcdffffff'When we compare by eye, we notice that the first three characters of needle and haystack match, but the fourth doesn't. Subsequently, we skip ahead when looking, because we know that the fourth character was an f, but needle[0] != 'f', so we can jump a few characters ahead. Specifically, we can start searching again from haystack[4:].
I suspect something similar can work in this problem, which would give it ultimately a runtime of O(nm), where n and m are the dimensions of the big matrix.
There was a problem hiding this comment.
this article is helping me a lot to understand KMP: http://jakeboxer.com/blog/2009/12/13/the-knuth-morris-pratt-algorithm-in-my-own-words/
There was a problem hiding this comment.
I haven't done thorough research yet, but it seems like this one also generalizes well to two-dimensional patern matching https://en.wikipedia.org/wiki/Rabin%E2%80%93Karp_algorithm
plz comment with thoughts, criticisms, anything! :)