Cosine matrix

[mishery]

Hi.

Is there a way to get a matrix of cosines between vectors?

I have a 518 column vectors of length 2160 and I was hoping to get a
518*518 matrix of the cosines between these vectors.

Basically I would like to the equivalent of cor(mydata) but for
cosines not correlations between the vectors. But of course,
cos(mydata) gives me the cosines of the individual values.

Can anyone help?

Thanks

- Mike

[Stefan Th. Gries]

This should do the trick:

# some random data
x<-matrix(rnorm(1000), ncol=10)

require(amap)
1-Dist(t(x), method="pearson")

HTH,
STG
--
Stefan Th. Gries
-----------------------------------------------
University of California, Santa Barbara
http://www.linguistics.ucsb.edu/faculty/stgries
-----------------------------------------------

[Ken Williams]

Or you can just use matrix algebra:

> x<-matrix(rnorm(40), ncol=5)
> x
[,1] [,2] [,3] [,4] [,5]
[1,] 0.1502795 -1.03924422 0.04422048 0.34105320 1.1524531
[2,] -1.7948242 -0.29985823 -0.70274315 -0.35564664 -2.2606329
[3,] -0.3682100 1.60098884 0.52262554 -0.05829224 -0.7012830
[4,] 0.5163998 -2.01919506 -1.12951358 1.55975383 1.0453210
[5,] -1.9530669 0.91230939 -0.22734366 1.13818820 0.3863500
[6,] 0.4367691 -0.09917415 0.89987130 1.11857546 0.1385893
[7,] 0.8961183 -0.25895326 -1.42279126 -0.34125845 0.9977009
[8,] -1.6253010 -0.43304923 -0.35767055 -0.68865311 1.1977955

> y <- sweep(x, 2, sqrt( colSums( x*x ) ), `/`) # Normalize
> t(y) %*% y
[,1] [,2] [,3] [,4] [,5]
[1,] 1.00000000 -0.2618860 0.08483994 0.07413309 0.3023032
[2,] -0.26188599 1.0000000 0.52274794 -0.30239278 -0.4309406
[3,] 0.08483994 0.5227479 1.00000000 -0.00876695 -0.2344688
[4,] 0.07413309 -0.3023928 -0.00876695 1.00000000 0.2923369
[5,] 0.30230321 -0.4309406 -0.23446878 0.29233689 1.0000000

-Ken

[mishery]

Thanks.

The matrix algebra version worked great, thank you. And I had not come
across "sweep" before, this will be very useful, so again thank you.

Can I just ask what the last backslash bit of function does?