| CPC G01C 21/3484 (2013.01) [G01C 21/3617 (2013.01)] | 18 Claims |
|
1. A navigation system comprising a server and a plurality of navigation client devices, wherein:
each navigation client device among the plurality of navigation client devices is configured to:
collect respective navigation history data representative of locations visited by the navigation client device;
use a respective training set of the respective navigation history data to train a respective local machine-learning model, the training including determining a respective set of updated parameters for the respective local machine-learning model, the parameters including coefficients of the respective local machine learning model; and
transmit, to the server, the respective set of updated parameters and
attribute data representative of at least one property of the respective training set;
the server is configured to:
receive, from each of the navigation client devices, the respective set of updated parameters and corresponding attribute data;
generate an aggregated set of parameters based on the respective sets of updated parameters, the generating including performing a weighted aggregation for which each of the respective sets of updated parameters is weighted in dependence on the corresponding attribute data; and
transmit data representative of the aggregated set of parameters to a first navigation client device among the plurality of navigation client devices; and the first navigation client device is configured to:
generate an updated respective local machine learning model, the generating including replacing parameters of the respective local machine-learning model with parameters from the aggregated set of parameters; and
provide, to a user, one or more outputs of the updated respective local machine learning model, the outputs being generated by processing data in the updated respective local machine learning model.
|