Content deleted Content added
No edit summary |
I propose that we stop trying to explain this algorithm in matrix form, and I make arguments in favor of this idea. |
||
Line 92:
I am generally new to editing at Wikipedia but tried to follow the guidelines I found to be relevant. If the content is not found suitable, contains errors or is otherwise unsuitable, I certainly welcome feedback, critique and any edits, including deletions, deemed necessary :)
[[User:BJJV|BJJV]] ([[User talk:BJJV|talk]]) 11:08, 26 May 2008 (UTC)
==Explaining this algorithm with Matrices==
It was a terrible idea to explain this algorithm in terms of matrices. Yes, it makes it so that you can elegantly describe the algorithm with a single formula, but it has many disadvantages, including:
#It makes the algorithm inaccessible to people that aren't comfortable with matrix math.
#It performs superfluous computations that would not be done in an efficient implementation.
#It hides context, and makes it easy to forget the meaning of columns and rows.
#The matrix math shown here is broken anyway. It shows row vectors where column vectors should be, and implies impossible calculations with matrices that aren't compatible for multiplication. When the smoothed values are computed, the symbol "x" is used, which implies cross-product, but straight scalar multiplication is used.
If we *really must* use matrix math to explain this algorithm, it needs to be cleaned up a lot. Frankly, I think the traditional explanation that uses probabilities and intuition is much better suited for an encyclopedic article. I propose that we rip out the matrix math completely. Also, I agree very much with the comment made earlier that this algorithm is so similar to the Viterbi algorithm that they should be explained together, and in a consistent manner.
| |||