Data efficient learning and rapid domain adaptation for wireless positioning and tracking

A device, comprising:

• a memory comprising computer-executable instructions;
• one or more processors configured to execute the computer executable instructions and cause the device to perform an operation comprising:
• receiving a first runtime record, wherein the first runtime record includes RF signal data collected in a physical space;
• and
• processing the first runtime record using a plurality of basis machine learning (ML) models to generate a plurality of inferences;
• aggregating the plurality of inferences to generate a plurality of coordinates;
• and
• outputting the plurality of coordinates, wherein the plurality of coordinates indicate a location of a physical element in a physical space.

A method, comprising:

• receiving a set of training records, each associated with a respective label comprising respective values for a plurality of output dimensions;
• decomposing each training record in the set of training records based on the plurality of output dimensions;
• and
• training a plurality of basis machine learning (ML) models based on the decomposed training records, comprising, for each respective output dimension of the plurality of output dimensions:
• determining, for a first training record, a respective dimension-specific value with respect to the respective output dimension based on a label of the first training record;
• and
• applying the respective dimension-specific value to generate a respective loss for a respective basis ML model of the plurality of basis ML models, wherein the respective basis ML model is specific to the respective output dimension.