Adaptive matrix distances aiming at optimum regression subspaces

MARC STRICKERT; AXEL J. SOTO; GUSTAVO E. VAZQUEZ

Bruges

Conferencia; European Symposium on Artificial Neural Networks, Computational Intelligence and Machine Learning - ESANN 2010; 2010

A  new  supervised  adaptive  metric  approach  is  introduced for  mapping  an  input  vector  space  to  a  plottable  low-dimensional  subspace  in  which  the  pairwise  distances  are  in  maximum  correlation  with distances of the associated target space.  The new formalism of multivariate subspace regression (MSR) is based on cost function optimization, and it allows assessing the relevance of input vector attributes.  An application to molecular descriptors in a chemical compound database is presented for targeting octanol-water partitioning properties.