On 31 Jan 2006, at 13:51, Theodore H. Smith wrote:
- speed related optimization of code, e.g. how using MemoryBlocks can
speed things up (yes, I'll do my homework about RegEx, Ed :)
You'll learn more about that by reading my plugin's website than
you'd get from your techniques.
- being endian aware (i.e. what you need to know to make your apps
work on Intel Macs or x-platform in general)
For example how my plugin has automatic UTF and endianness
detection for strings whose first 4 characters are ASCII (such as
XML).
Anything else you think I could talk about?
How to get levenshtein matrixes and smith-waterman matrixes to work
in as much a similar way as possible would be good. Seems like a
simple problem at first, but it's hard to work out the bugs. It's
hard to know where to start really, when all my standard approaches
for tackling problems, aren't bringing fruit on this particular one.
Maybe I'm not trying hard enough.
I think I realised the problem I'm having.
Actually the break I took from my work was useful. The problem wasn't
that I wasn't trying hard enough, but that I was focusing my efforts
on the wrong area. Trying too hard in other words.
In fact I need to figure out why does Smith-Waterman work the way it
does already. It's not that I couldn't implement it, that is easy,
but to understand the reasoning behind the implementation, that is an
order of magnitude harder.
For example Smith-Waterman uses a "gap penalty", while Levenshtein
doesn't. Now I need to figure out why does it need a gap penalty to
work properly, why not just say that an insert has the same scoring
effect as a replacement?
I think after I just understand Smith-Waterman thoroughly, I'll be
able to see how to merge the two algorithms, or just know if it is
impossible to merge, and thus I must duplicate my coding effort by
making two functions to do something very similar.
_______________________________________________
Unsubscribe or switch delivery mode:
<http://www.realsoftware.com/support/listmanager/>
Search the archives of this list here:
<http://support.realsoftware.com/listarchives/lists.html>
|