Adaptive pattern matcher with multidimensional predictors

A method, implemented using digital hardware and/or software logic, for predicting values for attributes-of-interest of objects, the method involving, for any given object in a series of incoming objects:

• storing a feature vector (FV) based on observable attributes of the given object, said observable attributes referred as predictor features (PFs);
• predicting a value for a latent attribute of the given object, with values for said latent features referred to as labels;
• ascertaining a value for the latent attribute of the given object, with such an ascertained value referred as the actual label of the given object;
• populating said actual label for the latent attribute in the FV of the given object, making it a labeled FV (LFV);
• and
• placing said LFV in one shift register (SR) among an array of label-specific SRs holding LFVs of objects accordant with their labels,
• wherein said predicting is done at least in part based on vector distance metrics between FV of the given object and the LFVs placed in their respective SR within the array of label-specific SRs.