Friday, November 26, 2004 18:50
The aim of this project is to create a series of new Max/MSP objects to facilitate the mapping of n to m parameters. These objects will be integrated in the MnM library for FTM.


linear mapping based on a training set of (input,output) vectors. The set can be either overdetermined or underdetermined.

The idea is to create a linear map (hyperplane) from an n-dimensional input space to an m-dimensional output space, just by specifying some input/output couples of examples. And the [matmap] object computes the map by least-square fitting.

(0.2,0.3,-0.1) -> (0.7,0.2,-0.9,1.9,2.5)
(0.6,0.7,0.8 ) -> (0.1,0.7,-0.1,0.3,0.9)
(0.5,0.2,-0.5) -> (0.9,0.1,-0.4,1.2,1.7)
(0.1,0.3,0.9 ) -> (0.5,0.3,-0.7,0.4,0.6)


Projection of an incoming vector on the principal components of a given training set.
To ease the hypridation between different data sets in the principal components space, all the coordinates are scaled by their singular values, and components with null singular values are automatically ignored.


objects for maxmsp 4.5 on win32
objects for maxmsp 4.5 on macosx