Showing posts with label numerical. Show all posts
Showing posts with label numerical. Show all posts

Wednesday, May 23, 2007

Invaders from The Fourth Dimension

So no updates for a couple days, but haven't been doing a whole lot of anything new until today.

Today I started rewriting my programs... again. Except this time, they're being implemented to work in four-dimensional space. How cool is that? It's not actually as simple as it sounds, though. Namely because, whereas in three dimensions every surface has only one straight line that is perpendicular to it (called the normal vector), in four dimensions those same surfaces have two lines that are perpendicular. Crazy, huh?

Friday, May 11, 2007

Week in Review: May 7-May 11

Okay, so I've missed a few days. Here's what went down. I started calculating the other kind of curvature, which is basically multiplying the same two values that were previously added to find the first kind. And I had to do that in two ways. The first way was proving to be horribly inaccurate (machine error popping up), but the second way (which calculates dead-on) is only valid for 3 dimensions. Once we go to four when only the first way will work. Quite the dilemma.

Tuesday, May 8, 2007

Symmetric, bitch

Today I rewrote my programs entirely from scratch... again. Except this time there was a useful point. Consider the shapes I was working with. A sphere, torus, parabloid... those are all round, aren't they? You can spin them all you want around the z-axis and it won't change how they look. Any surface like this is called rotationally symmetric and mathematically that means its curvature is dependent on only one variable instead of two.

So now my algorithms calculate the exact same stuff they did before, except they do it in a tiny fraction of the time it used to take, and the error shrinks much faster as well. So woot!

Thursday, May 3, 2007

Numerical Techniques: Part Three

Okay, so not a whole lot new accomplished today. Merged all the programs into three simpler programs, which makes it a lot easier to test new calculations. Though honestly, I have no idea whether these are coming out right or not :P. But, they do work out perfect for a sphere, and the only changes I make for any other surface are the formulas used to calculate the surfaces, so I'm supposing they work out right.

Wednesday, May 2, 2007

Numerical Techniques: Part Two

So today I expanded on the programs I made yesterday. Now it can handle a sphere, an ellipsoid, or torus, and ultimately finds what is called the "intrinsic curvature" of the object in terms of a particular distortion (currently involving just increasing the radius, essentially, but can be modified for any kind of distorting). I have no idea what this curvature actually represents, but apparently my programs calculate it right! That's always good.

Tuesday, May 1, 2007

Numerical Techniques

Today I began implementing some simple numerical differentiation and integration routines, applying them to some techniques found in differential geometry.

In the span of nearly three hours, I wrote a series of five programs. These programs must be run sequentially, as each uses the data of the previous program (except, of course, for the first which generates the data).

Five programs and three hours and I can successfully calculate the area of a sphere with radius 1.

Okay, so maybe I'm demeaning this process a little. My initial application was a unit sphere. But the first program can be modified to generate data for any kind of surface, and the remaining programs can find the area of said surface with little-to-no modification.