| ||||||||||||||||||
| Current Server Time - 4:39:51 AM |         |
GameDev.Net
Forums  Math
and Physics 4d vectors and matrices - which operators
are defined?Send Topic To a Friend | View Forum FAQ | Watch this thread |
   |
  Milamber12345 Member since: 9/18/2003
From: New South Wales, Australia   |
 Posted - 18 September 2003 8:51:37 AM
|
| I am writing some c++ classes to handle vector,
matrix, and quaternion math, but I can't work out how to implement some of
the operators, eg. should scalar addition be defined for a 4d vector? If
so, I don't think it should apply to the w coordinate. Also, what is the
definition of the cross product of 2 4d vectors? Similar questions for the
4d matrices. |
| Kentaro Member since: 11/28/2002
From: Canada |
| Posted - 18 September 2003 10:35:29 AM
|
| As far as I can tell, the cross product of two vectors
is only defined in 3 dimensions. I've never even heard of adding a scalar
to a vector. The basic vector operations apply in n dimensions. v + u = (v1 + u1, ..., vn + un) v (dot) u = v1(u1) + ... + vn(un) You can also prove some other properties in n dimensions (distributivity and whatnot). As for matrices, for two matrices A and B that have dimensions m x n, their sum is an m x n matrix with each entry being the sum of that entry in A and that entry in B. A + B[i, j] = A[i,j] + B[i, j] Determinants... I'm not even sure why you're asking this. Most operations that I know of apply in any dimension! (Well, I have no clue about quaternions) Over the centuries, mankind has tried many ways of combating the forces of evil...prayer, fasting, good works and so on. Up until Doom, no one seemed to have thought about the double-barrel shotgun. Eat leaden death, demon... -- Terry Pratchett |
| Anonymous Poster |
| Posted - 18 September 2003 2:19:56 PM
|
| What you use will depend on whether these are true 4d
vectors, or 3d vectors in homogeneous coordinates. I.e., for a true 4d
vector, the cross product is undefined, while it is for a homogenous 3d
vector. I.e. given two 3d vectors in homogenous coordinates, V1,w1, and V2,w2. The cross product would be defined as (1/w1)*V1 x (1/w2)V2,1. I suggest you read up on the theory of homogenous coordinates and fully understand the theory before getting too far in code. |
| alvaro Member since: 3/7/2002
From: USA |
| Posted - 18 September 2003 2:22:25 PM
|
| By the definition of vector space, sum of vectors and
product of a number (a scalar) and a vector are always defined. nxn matrices are vectors too (in a space with n^2 dimensions), so those two operations apply. Adding a number to a vector is not meaningful. The cross product is only defined in dimension 3. There is something called external product that is defined in other dimensions, but the result is not a regular vector and it's probably not very interesting for your purposes. Dot product is defined in R^n as Kentaro did. A very important operation is multiplying a matrix by a [column] vector, with the result begin a [column] vector. There are several other things that you can do, but you can add them as you need them. For instance, there is no point in implementing the double ratio of four points if you are never going to use it. [edited by - Alvaro on September 18, 2003 2:23:23 PM] |
| Milamber12345 Member since: 9/18/2003
From: New South Wales, Australia |
| Posted - 18 September 2003 5:36:17 PM
|
| Thanks for your replies everyone, and just to clarify,
the 4d vectors and matrices are for handling 3d translations, so, yes, I'd
say homogeneous coordinates. I'm working my way through a book called
"Mathematics for 3d game programming and computer graphics". The first 3
chapters - vectors, matrices, and transforms - are intended for revision,
though, and are a little brief for learning the basic concepts from. Any
other pointers for how operatoors are defined for these would be useful.
BTW, how would you implement the determinant and inverse of a 4x4 matrix
for homogeneous coords? They are defined for n dimensions, but is the w
coordiate treated normally? Also, when making the transpose of a 4x4
matrix, do you shift the translation vector as well? If not, do you change
it in any way? Thx in advance for any further clarification. |
| Sneftel Moderator
Member since: 7/7/2001 From:
|
| Posted - 18 September 2003 5:42:08 PM
|
| Determinants, inverses, and transpositions all treat
the w component just like all the rest of 'em. Note, however, that it is
unlikely you will ever take the transposition of a matrix with anything
other than (0,0,0,1) in the fourth column--consider the possibility of a
logic error if this ever happens. How appropriate. You fight like a cow. |
| python_regious Member since: 2/28/2001
From: Somewhere in the vauge vicinity of Betelgeuse
|
| Posted - 18 September 2003 6:06:13 PM
|
| Hmm... Couldn't the cross product in 4 dimensions just
be the expansion of the equivilent 4x4 determinant? That makes more sense
than just "it's undefined"... I mean, in two dimensions it could be considered that:
would be correct, and if you expand it you have the function for finding a normal vector to another vector in 2D space. Now, using 3 4-dimensional vectors, couldn't a normal to a 3D space be found? You have to remember that you're unique, just like everybody else. 
|
| Anonymous Poster |
| Posted - 18 September 2003 8:31:54 PM
|
| If one defines the cross product by these
properties: 1) Works on two vectors, say A and B 2) Produces a vector C, perpendicular to both A and B 3) The length of C is |A||B|sin theta, where theta is the angle in-between A and B. Then, the cross product can only be defined in 3, and 7 dimensions. See: http://www.helsinki.fi/~lounesto/cross.product |
| Kentaro Member since: 11/28/2002
From: Canada |
| Posted - 18 September 2003 11:54:27 PM
|
| The link you give is actually even more general than
that. It defines the cross product of k vectors in n dimensions: k = 2, n = 3 or 7 k = 3, n = 8 ... and some others. That's very interesting. I was having a debate with someone in my multi-variable calc course who said he could define the cross product of two vectors in any dimension so I should show this to him. I intuitively thought that the cross product of two vectors could not be defined in 4-space. Thankfully, the dot product is a lot easier to define and I seem to have gotten it right. =) |
| python_regious Member since: 2/28/2001
From: Somewhere in the vauge vicinity of Betelgeuse
|
| Posted - 19 September 2003 11:51:26 AM
|
quote: Yes, this is what I thought, so, using the ( n, n-1 ) formulae, it can be seen that the cross product is valid for 4 dimensional space. I presume this calculates the vector perpendicular to the volume defined by the 3 4D vectors... You have to remember that you're unique, just like everybody else.
|
| All times are ET (US)
|
|
Administrative Options:
About Us | Advertise on GameDev.net | Write for us
© 1999-2004
Gamedev.net, LLC. All rights reserved. Terms of Use
Privacy
Policy
Comments? Questions? Feedback? Click here!
GameDev.netTM, the GameDev.net logo, and
GDNetTM are trademarks of GameDev.net,
LLC
