[<< | Prev | Index | Next | >>]

Thursday, August 15, 2002

Just Another Notch in the Wheels of Time



My net has been down 95% of the time for the last week. And to top it off, now that some email has finally trickled in during one of the breif up-times, Eudora is crashing again (immediately upon launch!).

Am I the only person who has Cast Away (the movie) envy?


I got two hang-up messages, and a third asking for a Mrs. Domingez to return the call. I have been getting calls for the Domingez household with some regularity for a few years now, and whenever I answer the phone the people calling always confirm it is my number they are trying to dial.

So this guy who called my machine three times called back the next day, and when I told him that this happens all the time, he told me I should call the phone company because sometimes they get their wires crossed and clearly they need to fix this.

No, I'm not joking, and neither was he.

And then he abruptly and rudely hung up on me since evidently I was not immediately seeking his services as a screen door repairman, as Mrs. Domingez apparently is, and heaven knows there's no reason not to be an asshole to anyone who's favor you don't need at the moment.

But he left his number on my machine the other day, so I called him back, waited patiently while he finished dialing after failing to check for a dial tone, and continued our nice little conversation. My goal, if possible, was to get him to play me Mrs. Domingez's message so I could possibly figure out her number myself for future reference, and to attempt to get it into this guy's pea brain that it is not the phone company's wires that are crossed, but Mrs. Domingez's, and that if he continues to dial my phone number, he will continue to reach me.

I confirmed again with him that the number he was trying to call was, in fact, my phone number, which I made clear to him I have had for many many years. "And, therefor, clearly the problem does not lie with the phone company..." "Yes it does!" he replied emphatically, then hung up on me.

I know, I know -- screen door repairman, what do I expect? Still, it's not about IQ, but attitude.

Yet, I have no doubt he'll have more kids than me. Funny the things the universe cares about, or doesn't.


One of the first applications I wanted to create with my new AI tech is equation recognition (parsing hand-written equations). It's something my tech is well suited to, and which nobody else seems to have made any headway in; though every time I mention the idea to anyone they remind me that there's no great demand for such a product so I'm not apt to make any money at it.

So today I read this article talking about Microsoft giving $2.3 million to the University of Waterloo, for which (along with the rider of requiring future students to take a C# [Microsoft's new programming language] course in order to graduate), "the aim of the research project is to develop equation recognition ..."

Great, just what I need--a 800 lb gorilla pissing in my yard.

I think I'll go back to working on my mechanical simulator.


Speaking of which, does anybody know a fast way to invert symmetric singular matrices? Yeah yeah, "you can't invert singular matrices". Not true -- there should be an inverse which is at least valid for vectors within the range of the matrix (the subspace spanned by its rows or columns). I came up with one, which works fairly well but still doesn't seem quite right to me since the result is not symmetric and my intuition says there should be a symmetric solution.

Garrett* suggested iterative methods for dividing a vector by a matrix, so I did some research and again found the standard methods don't converge for singular matrices, or the ones that do aren't symmetric (even less so than my inversion method; and for my purposes the symmetry is important). So once again I came up with my own, which is symmetric, converges for singular matrices, and is faster than the standard assymetric solutions (which require a partial matrix inverse before beginning the iterations), but still it runs about half the speed of just using the matrix inverse, and its convergence time is less bounded in general, so I'm back to looking for a quick, symmetric, singular matrix inversion method. "Dear Abby..."

What's it all for? Mechanical simulation of robots, of course... :)

[<< | Prev | Index | Next | >>]


Simon Funk / simonfunk@gmail.com