SDBOX Bounding box classifier P=SDBOX(DATA) INPUT DATA data set OUTPUT P bounding box pipeline OPTIONS 'bounds',[min max] - define minimum and maximum (scalar) bounds applicable to all features. 'min',M - define minimum value (M may be vector with min per feature) 'max',M2 - define minimum value (M may be vector with min per feature) 'reject',P - reject P-percentile lowest and highest values DESCRIPTION SDBOX defines a bounding box classifier for the entire data set DATA. By default minimum and maximum values for each feature are used. Alternatively, we may provide percentile to be rejected using 'reject',P option. If scalar P is provided, P/2 percentile is rejected from bottom and P/2 from the top distribution of each feature. P may be 2-component vector directly defining low and high-percentile. The 'bounds' option allows user to set specific min and max scalar values used for all features. This is meaningful when all features have identical bounds. Alternatively, min and max bounds per-feature may be set using 'min' and 'max' options, resp. SDBOX uses 'inside' and 'outside' decisions by default. EXAMPLE: Protect data in a PCA subspace in all directions: >> p=sdpca*sdbox; pbox=data*p Make sure data is between 0 and 5000: >> p=sdbox(data,'bounds',[0 5000]); SEE ALSO SDPCA, SDRELAB