Quote from last Wednesday

[kirisutogomen]

"Every time you use a matrix in a proof, a puppy dies."

Comments

[crs.livejournal.com]

What happens if you use 4000 lines of code?

[rifmeister.livejournal.com]

Lordy. I kill like twenty puppies every week.
I rarely do a proof that *doesn't* use matrices.

[arcanology.livejournal.com]

Your crimes against puppies are unforgivable.

I shall shun you now.

Yeah...

[mjperson.livejournal.com]

As soon as I read the post, I thought, "Man, we'll have to call rifmeister the Puppyslayer..."

Good that you can admit it.

Dead Puppae

[kirisutogomen.livejournal.com]

The statement was made by a dude who used to work in speech recognition, and he assures me that using matrices in a proof should be done only in desperation.

Re: Dead Puppae

[rifmeister.livejournal.com]

I don't understand. Maybe he means something different than I do by "using matrices in a proof." More likely, he means something different than I do by "a puppy dies."

Re: Dead Puppae

[kirisutogomen.livejournal.com]

We're making a distinction between a linear transformation with no specified coordinate system and a matrix, tied to a specific basis.

Re: Dead Puppae

[rifmeister.livejournal.com]

That's a little abstract for me. Usually, I start with a bunch of data points, and I put them in a matrix called X. Then I form X^t X or some generalization thereof and take an eigendecomposition or an SVD or a Cholesky or whatever. I don't think I'm talking about abstract linear transformations. I'm talking about data. And killing puppies.

What's this guy teaching, anyways?

Re: Dead Puppae

[kirisutogomen.livejournal.com]

There's your problem. Proofs don't have "data points." Data is for scientists. Or maybe engineers, but who cares what they do?

I'm taking Classical Geometry.

[remcat.livejournal.com]

No worries.

There are plenty of horrible little puppies to go around :).