SDPROX Proximity representationComputate proximities to prototypes, return data set D=sdprox(DATA,PROTO) D=sdprox(DATA,PROTO,TYPE,PAR) Creating proximity pipeline for computing proximities later PP=SDPROX(PROTO,TYPE,PAR) Create proximity data set by providing the raw data matrix. D=SDPROX(DATA,PROTO,'matrix',D_raw) INPUT DATA SDDATA set or data matrix with samples to be used for direct proximity computation PROTO SDDATA or data matrix with prototype samples TYPE proximity type SQEUCL squared Euclidean distance (default) EUCL Euclidean distance LINEAR linear form (inner product) POLY polynomial kernel (PAR degree, def=3) RBF radial basis kernel (PAR sigma, def=1.0) SAM Spectral Angle Mapper ASAM ACOS of Spectral Angle Mapper KOL Kolmogorov distance (assummes that feature vectors sum-to-one, use SDNORM) MATCH Matching distance (sum of differences between cum.distr.func, assumes features sum-to-one) PAR proximity parameter (optional) D_raw Raw proximity matrix (num.of samples in DATA x num.of samples in PROTO) OUTPUT PP pipeline object D SDDATA set with distances from samples in DATA to PROTO DESCRIPTION SDPROX computes proximities to a set of prototypes. If DATA and PROTO are provided, SDPROX returns SDDATA object with proximity values. If only PROTO is given, SDPROX returns the pipline PP which may be applied to any new DATA to compute proximities. The raw matrix of proximities D_raw may be provided which is computed by custom code. SDPROX then creates a data set D which contains sample properties of DATA and data matrix D_raw. It copies sample properties of PROTO making them feature properties in D (adding 'proto_' prefix). EXAMPLES >> D=sdprox(data,proto) % compute squared Euclidean distances >> D=sdprox(data,proto,'SAM') % compute Spectral Angle Mapper distances >> pp=sdprox(proto,'eucl'); D=data*pp % create and apply proximity pipeline